abstract
stringlengths 383
2.62k
| TLDR
stringlengths 4
250
|
|---|---|
Markov logic networks (MLNs) are powerful models for symbolic reasoning, which combine probabilistic modeling with relational logic. Inference algorithms for MLNs often perform at the level of propositional logic or require building a first-order probabilistic graph, and the computational efficiency remains a challenge. The mean-field algorithm generalizes message passing for approximate inference in many intractable probabilistic graphical models, but in MLNs it still suffers from the high-order dependencies among the massive groundings, resulting in time complexity exponential in both the length and the arity of logic rules. We propose a novel method, LogicMP, to simplify the logic message passing especially. In most practical cases, it can reduce the complexity significantly to polynomial for the formulae in conjunctive normal form (CNF). We exploit the property of CNF logic rules to sidestep the expectation computation of high-order dependency, and then formulate the logic message passing by Einstein summation to facilitate parallel computation, which can be optimized by sequentially contracting the rule arguments. With LogicMP, we achieve evident improvements on several reasoning benchmark datasets in both performance and efficiency over competitor methods. Specifically, the AUC-PR of the UW-CSE and Cora datasets is improved by more than 11\% absolutely and the speed is about ten times faster.
| This paper proposes a method to perform mean-field iteration of MLN efficiently via a novel neural network. |
Hopfield networks are artificial neural networks which store memory patterns on the states of their neurons by choosing recurrent connection weights and update rules such that the energy landscape of the network forms attractors around the memories. How many stable, sufficiently-attracting memory patterns can we store in such a network using $N$ neurons? The answer depends on the choice of weights and update rule. Inspired by setwise connectivity in biology, we extend Hopfield networks by adding setwise connections and embedding these connections in a simplicial complex. Simplicial complexes are higher dimensional analogues of graphs which naturally represent collections of pairwise and setwise relationships. We show that our simplicial Hopfield networks increase memory storage capacity. Surprisingly, even when connections are limited to a small random subset of equivalent size to an all-pairwise network, our networks still outperform their pairwise counterparts. Such scenarios include non-trivial simplicial topology. We also test analogous modern continuous Hopfield networks, offering a potentially promising avenue for improving the attention mechanism in Transformer models. | Without increasing the number of parameters, we improve the memory capacity of Hopfield networks by adding setwise connections embedded in a simplicial complex. |
The Strong Lottery Ticket Hypothesis (SLTH) stipulates the existence of a subnetwork within a sufficiently overparameterized (dense) neural network that---when initialized randomly and without any training---achieves the accuracy of a fully trained target network. Recent works by Da Cunha et. al 2022, Burkholz 2022 demonstrate that the SLTH can be extended to translation equivariant networks---i.e. CNNs---with the same level of overparametrization as needed for the SLTs in dense networks. However, modern neural networks are capable of incorporating more than just translation symmetry, and developing general equivariant architectures such as rotation and permutation has been a powerful design principle. In this paper, we generalize the SLTH to functions that preserve the action of the group $G$---i.e. $G$-equivariant network---and prove, with high probability, that one can approximate any $G$-equivariant network of fixed width and depth by pruning a randomly initialized overparametrized $G$-equivariant network to a $G$-equivariant subnetwork. We further prove that our prescribed overparametrization scheme is optimal and provide a lower bound on the number of effective parameters as a function of the error tolerance. We develop our theory for a large range of groups, including subgroups of the Euclidean $\text{E}(2)$ and Symmetric group $G \leq \mathcal{S}_n$---allowing us to find SLTs for MLPs, CNNs, $\text{E}(2)$-steerable CNNs, and permutation equivariant networks as specific instantiations of our unified framework. Empirically, we verify our theory by pruning overparametrized $\text{E}(2)$-steerable CNNs, $k$-order GNNs, and message passing GNNs to match the performance of trained target networks. | We extend the strong lottery ticket hypothesis to Equivariant Networks and show optimal pruning strategies in theory and practice for Steerable CNNs, Higher Order GNNs, and Message Passing GNNs. |
A key challenge towards differentially private machine learning is balancing the trade-off between privacy and utility.
A recent line of work has demonstrated that leveraging \emph{public data samples} can enhance the utility of DP-trained models (for the same privacy guarantees).
In this work, we show that public data can be used to improve utility in DP models significantly more than shown in recent works.
Towards this end, we introduce a modified DP-SGD algorithm that leverages public data during its training process.
Our technique uses public data in two complementary ways: (1) it uses generative models trained on public data to produce synthetic data that is effectively embedded in multiple steps of the training pipeline; (2) it uses a new gradient clipping mechanism (required for achieving differential privacy) which changes the \emph{origin} of gradient vectors using information inferred from available public and synthesized data.
Our experimental results demonstrate the effectiveness of our approach in improving the state-of-the-art in differentially private machine learning across multiple datasets, network architectures, and application domains.
Notably, we achieve a $75\%$ accuracy on CIFAR10 when using only $2,000$ public images; this is \emph{significantly higher} than the state-of-the-art which is $68\%$ for DP-SGD with the privacy budget of $\varepsilon=2,\delta=10^{-5}$ (given the same number of public data points). | improving the effect of public data in DP-SGD and improving the accuracy significantly |
Machine learning force fields (MLFF) have been proposed to accelerate molecular dynamics (MD) simulation, which finds widespread applications in chemistry and biomedical research. Even for the most data-efficient MLFF models, reaching chemical accuracy can require hundreds of frames of force and energy labels generated by expensive quantum mechanical algorithms, which may scale as $O(n^3)$ to $O(n^7)$, with $n$ being the number of basis functions used and typically proportional to the number of atoms.
To address this issue, we propose a multi-stage computational framework -- ASTEROID, which enjoys low training data generation cost without significantly sacrificing MLFFs' accuracy. Specifically, ASTEROID leverages a combination of both large cheap inaccurate data and small expensive accurate data. The motivation behind ASTEROID is that inaccurate data, though incurring large bias, can help capture the sophisticated structures of the underlying force field. Therefore, we first train a MLFF model on a large amount of inaccurate training data, employing a bias-aware loss function to prevent the model from overfitting the potential bias of the inaccurate training data. We then fine-tune the obtained model using a small amount of accurate training data, which preserves the knowledge learned from the inaccurate training data while significantly improving the model's accuracy. Moreover, we propose a variant of ASTEROID based on score matching for the setting where the inaccurate training data are unlabelled. Extensive experiments on MD simulation datasets show that ASTEROID can significantly reduce data generation costs while improving the accuracy of MLFFs. | We propose ASTEROID, a computational framework to reduce the data generation cost of training machine learning force fields. |
Prediction of a varying number of ordered clusters from sets of any cardinality is a challenging task for neural networks, combining elements of set representation, clustering and learning to order. This task arises in many diverse areas, ranging from medical triage, through multi-channel signal analysis for petroleum exploration to product catalog structure prediction. This paper focuses on the latter, which exemplifies a number of challenges inherent to adaptive ordered clustering, referred to further as the eponymous Catalog Problem. These include learning variable cluster constraints, exhibiting relational reasoning and managing combinatorial complexity. Despite progress in both neural clustering and set-to-sequence methods, no joint, fully differentiable model exists to-date. We develop such a modular architecture, referred to further as Neural Ordered Clusters (NOC), enhance it with a specific mechanism for learning cluster-level cardinality constraints, and provide a robust comparison of its performance in relation to alternative models. We test our method on three datasets, including synthetic catalog structures and PROCAT, a dataset of real-world catalogs consisting of over 1.5M products, achieving state-of-the-art results on a new, more challenging formulation of the underlying problem, which has not been addressed before. Additionally, we examine the network's ability to learn higher-order interactions and investigate its capacity to learn both compositional and structural rulesets. | A neural method for predicting an adaptive number of diverse, ordered clusters from any set is introduced and tested on synthetic and real-world datasets, demonstrating top performance on new and harder formulation of the PROCAT challenge. |
Learning prediction models that generalize to related domains is one of the most fundamental challenges in artificial intelligence. There exists a growing literature that argues for learning invariant associations using data from multiple source domains. However, whether invariant predictors generalize to a given target domain depends crucially on the assumed structural changes between domains. Using the perspective of transportability theory, we show that invariance learning, and the settings in which invariant predictors are optimal in terms of worst-case losses, is a special case of a more general partial transportability task. Specifically, the partial transportability task seeks to identify / bound a conditional expectation $\mathbb E_{P_{\pi^*}}[y\mid\mathbf x]$ in an unseen domain $\pi^*$ using knowledge of qualitative changes across domains in the form of causal graphs and data from source domains $\pi^1,\dots,\pi^k$. We show that solutions to this problem have a much wider generalization guarantee that subsumes those of invariance learning and other robust optimization methods that are inspired by causality. For computations in practice, we develop an algorithm that provably provides tight bounds asymptotically in the number of data samples from source domains for any partial transportability problem with discrete observables and illustrate its use on synthetic datasets. | This paper investigates the problem of domain generalization from the perspective of transportability theory. We propose the task of partial transportability and provide solutions that highlight new contrasts with "invariance learning" methods. |
Labor economists regularly analyze employment data by fitting predictive models to small, carefully constructed longitudinal survey datasets. Although modern machine learning methods offer promise for such problems, these survey datasets are too small to take advantage of them. In recent years large datasets of online resumes have also become available, providing data about the career trajectories of millions of individuals. However, standard econometric models cannot take advantage of their scale or incorporate them into the analysis of survey data. To this end we develop CAREER, a transformer-based model that uses transfer learning to learn representations of job sequences. CAREER is first fit to large, passively-collected resume data and then fine-tuned to smaller, better-curated datasets for economic inferences. We fit CAREER to a dataset of 24 million job sequences from resumes, and fine-tune its representations on longitudinal survey datasets. We find that CAREER forms accurate predictions of job sequences, achieving state-of-the-art predictive performance on three widely-used economics datasets. We further find that CAREER can be used to form good predictions of other downstream variables; incorporating CAREER into a wage model provides better predictions than the econometric models currently in use. | We develop CAREER, a transformer-based model of job sequence data, which is pretrained on large resume datasets and fine-tuned to survey datasets widely used by labor economists. |
In this paper we propose a new Deep Learning (DL) approach for message classification. Our method
is based on the state-of-the-art Natural Language Processing (NLP) building blocks, combined
with a novel technique for infusing the meta-data input that is typically available in messages
such as the sender information, timestamps, attached image, audio, affiliations, and more. As we
demonstrate throughout the paper, going beyond the mere text by leveraging all available channels
in the message, could yield an improved representation and higher classification accuracy. To
achieve message representation, each type of input is processed in a dedicated block in the neural
network architecture that is suitable for the data type. Such an implementation enables training all
blocks together simultaneously, and forming cross channels features in the network. We show in the
Experiments Section that in some cases, message’s meta-data holds an additional information that
cannot be extracted just from the text, and when using this information we achieve better performance.
Furthermore, we demonstrate that our multi-modality block approach outperforms other approaches
for injecting the meta data to the the text classifier. | We propose a deep neural network based on blocks for message classification using meta-data inputs |
The deployment of Deep Learning (DL) models is still precluded in those contexts where the amount of supervised data is limited. To answer this issue, active learning strategies aim at minimizing the amount of labelled data required to train a DL model. Most active strategies are based on uncertain sample selection, and even often restricted to samples lying close to the decision boundary. These techniques are theoretically sound, but an understanding of the selected samples based on their content is not straightforward, further driving non-experts to consider DL as a black-box. For the first time, here we propose a different approach, taking into consideration common domain-knowledge and enabling non-expert users to train a model with fewer samples. In our Knowledge-driven Active Learning (KAL) framework, rule-based knowledge is converted into logic constraints and their violation is checked as a natural guide for sample selection. We show that even simple relationships among data and output classes offer a way to spot predictions for which the model need supervision. The proposed approach (i) outperforms many active learning strategies in terms of average F1 score, particularly in those contexts where domain knowledge is rich. Furthermore, we empirically demonstrate that (ii) KAL discovers data distribution lying far from the initial training data unlike uncertainty-based strategies, (iii) it ensures domain experts that the provided knowledge is respected by the model on test data, and (iv) it can be employed even when domain-knowledge is not available by coupling it with a XAI technique. Finally, we also show that KAL is also suitable for object recognition tasks and, its computational demand is low, unlike many recent active learning strategies. | Exploiting simple domain-knowledge within an active learning strategy to minimize the amount of labelled data and whiten the labelling process. |
Distributional shift, or the mismatch between training and deployment data, is a significant obstacle to the usage of machine learning in high-stakes industrial applications, such as autonomous driving and medicine. This creates a need to be able to assess how robustly ML models generalize as well as the quality of their uncertainty estimates. Standard ML datasets do not allow these properties to be assessed, as the training, validation and test data are often identically distributed. Recently, a range of dedicated benchmarks have appeared, featuring both distributionally matched and shifted data. The Shifts dataset stands out in terms of the diversity of tasks and data modalities it features. Unlike most benchmarks, which are dominated by 2D image data, Shifts contains tabular weather forecasting, machine translation, and vehicle motion prediction tasks. This enables models to be assessed on a diverse set of industrial-scale tasks and either universal or directly applicable task-specific conclusions to be reached. In this paper, we extend the Shifts Dataset with two datasets sourced from industrial, high-risk applications of high societal importance. Specifically, we consider the tasks of segmentation of white matter Multiple Sclerosis lesions in 3D magnetic resonance brain images and the estimation of power consumption in marine cargo vessels. Both tasks feature ubiquitous distributional shifts and strict safety requirements due to the high cost of errors. These new datasets will allow researchers to explore robust generalization and uncertainty estimation in new situations. This work provides a description of the dataset and baseline results for both tasks. | We introduce two new datasets into the Shifts Benchmark for assessing robustness and uncertainty - Multiple Sclerosis Lesion Segmentation in MRI images and Cargo Vessel Power Consumption Prediction |
We consider the problem of making AI agents that collaborate well with humans in partially observable fully cooperative environments given datasets of human behavior. Inspired by piKL, a human-data-regularized search method that improves upon a behavioral cloning policy without diverging far away from it, we develop a three-step algorithm that achieve strong performance in coordinating with real humans in the Hanabi benchmark. We first use a regularized search algorithm and behavioral cloning to produce a better human model that captures diverse skill levels. Then, we integrate the policy regularization idea into reinforcement learning to train a human-like best response to the human model. Finally, we apply regularized search on top of the best response policy at test time to handle out-of-distribution challenges when playing with humans. We evaluate our method in two large scale experiments with humans. First, we show that our method outperforms experts when playing with a group of diverse human players in ad-hoc teams. Second, we show that our method beats a vanilla best response to behavioral cloning baseline by having experts play repeatedly with the two agents. | a new method for human-AI collaboration based on human regularized search, imitation learning and RL, tested with large scale human experiments. |
While unsupervised skill discovery has shown promise in autonomously acquiring behavioral primitives, there is still a large methodological disconnect between task-agnostic skill pretraining and downstream, task-aware finetuning. We present Intrinsic Reward Matching (IRM), which unifies these two phases of learning via the $\textit{skill discriminator}$, a pretraining model component often discarded during finetuning. Conventional approaches finetune pretrained agents directly at the policy level, often relying on expensive environment rollouts to empirically determine the optimal skill. However, often the most concise yet complete description of a task is the reward function itself, and skill learning methods learn an $\textit{intrinsic}$ reward function via the discriminator that corresponds to the skill policy. We propose to leverage the skill discriminator to $\textit{match}$ the intrinsic and downstream task rewards and determine the optimal skill for an unseen task without environment samples, consequently finetuning with greater sample-efficiency. Furthermore, we generalize IRM to sequence skills and solve more complex, long-horizon tasks. We demonstrate that IRM is competitive with previous skill selection methods on the Unsupervised Reinforcement Learning Benchmark and enables us to utilize pretrained skills far more effectively on challenging tabletop manipulation tasks. | We unify unsupervised RL skill pretraining and downstream finetuning phases of learning by leveraging the skill discriminator as a task specifier. |
Normalizing flows are tractable density models that can approximate complicated target distributions, e.g. Boltzmann distributions of physical systems. However, current methods for training flows either suffer from mode-seeking behavior, use samples from the target generated beforehand by expensive MCMC methods, or use stochastic losses that have high variance. To avoid these problems, we augment flows with annealed importance sampling (AIS) and minimize the mass-covering $\alpha$-divergence with $\alpha=2$, which minimizes importance weight variance. Our method, Flow AIS Bootstrap (FAB), uses AIS to generate samples in regions where the flow is a poor approximation of the target, facilitating the discovery of new modes. We apply FAB to multimodal targets and show that we can approximate them very accurately where previous methods fail. To the best of our knowledge, we are the first to learn the Boltzmann distribution of the alanine dipeptide molecule using only the unnormalized target density, without access to samples generated via Molecular Dynamics (MD) simulations: FAB produces better results than training via maximum likelihood on MD samples while using 100 times fewer target evaluations. After reweighting the samples, we obtain unbiased histograms of dihedral angles that are almost identical to the ground truth. | We train normalizing flows to fit multi-modal target distributions by generating samples where the flow is a poor approximation of the target using an annealed importance sampling bootstrap procedure. |
Recurrent neural networks (RNNs) have brought a lot of advancements in sequence labeling tasks and sequence data. However, their effectiveness is limited when the observations in the sequence are irregularly sampled, where the observations arrive at irregular time intervals. To address this, continuous time variants of the RNNs were introduced based on neural ordinary differential equations (NODE). They learn a better representation of the data using the continuous transformation of hidden states over time, taking into account the time interval between the observations. However, they are still limited in their capability as they use the discrete transformations and discrete number of layers (depth) over an input in the sequence to produce the output observation. We intend to address this limitation by proposing RNNs based on differential equations which model continuous transformations over depth and time to predict an output for a given input in the sequence. Specifically, we propose continuous depth recurrent neural differential equations (CDR-NDE) which generalizes RNN models by continuously evolving the hidden states in both the temporal and depth dimensions. CDR-NDE considers two separate differential equations over each of these dimensions and models the evolution in the temporal and depth directions alternatively. We also propose the CDR-NDE-heat model based on partial differential equations which treats the computation of hidden states as solving a heat equation over time. We demonstrate the effectiveness of the proposed models by comparing against the state-of-the-art RNN models on real world sequence modeling problems and data sets. | Proposing novel RNN models based on differential equations that continuously transform hidden states in both temporal and depth dimensions. |
We consider the task of generating realistic 3D shapes, which is useful for a variety of applications such as automatic scene generation and physical simulation. Compared to other 3D representations like voxels and point clouds, meshes are more desirable in practice, because (1) they enable easy and arbitrary manipulation of shapes for relighting and simulation, and (2) they can fully leverage the power of modern graphics pipelines which are mostly optimized for meshes. Previous scalable methods for generating meshes typically rely on sub-optimal post-processing, and they tend to produce overly-smooth or noisy surfaces without fine-grained geometric details. To overcome these shortcomings, we take advantage of the graph structure of meshes and use a simple yet very effective generative modeling method to generate 3D meshes. Specifically, we represent meshes with deformable tetrahedral grids, and then train a diffusion model on this direct parameterization. We demonstrate the effectiveness of our model on multiple generative tasks. | Diffusion model on 3D meshes of arbitrary topology by direct parametrizing meshes with tetrahedral grids |
In this paper, we study the problem of 3D scene geometry decomposition and manipulation from 2D views. By leveraging the recent implicit neural representation techniques, particularly the appealing neural radiance fields, we introduce an object field component to learn unique codes for all individual objects in 3D space only from 2D supervision. The key to this component is a series of carefully designed loss functions to enable every 3D point, especially in non-occupied space, to be effectively optimized even without 3D labels. In addition, we introduce an inverse query algorithm to freely manipulate any specified 3D object shape in the learned scene representation. Notably, our manipulation algorithm can explicitly tackle key issues such as object collisions and visual occlusions. Our method, called DM-NeRF, is among the first to simultaneously reconstruct, decompose, manipulate and render complex 3D scenes in a single pipeline. Extensive experiments on three datasets clearly show that our method can accurately decompose all 3D objects from 2D views, allowing any interested object to be freely manipulated in 3D space such as translation, rotation, size adjustment, and deformation. | In this paper, we study the problem of 3D scene geometry decomposition and manipulation from 2D views. |
Existing Deep Reinforcement Learning (DRL) algorithms suffer from sample inefficiency. Generally, episodic control-based approaches are solutions that leverage highly rewarded past experiences to improve sample efficiency of DRL algorithms. However, previous episodic control-based approaches fail to utilize the latent information from the historical behaviors (\eg, state transitions, topological similarities, \etc) and lack scalability during DRL training. This work introduces Neural Episodic Control with State Abstraction (NECSA), a simple but effective state abstraction-based episodic control containing a more comprehensive episodic memory, a novel state evaluation, and a multi-step state analysis. We evaluate our approach to the MuJoCo and Atari tasks in OpenAI gym domains. The experimental results indicate that NECSA achieves higher sample efficiency than the state-of-the-art episodic control-based approaches. Our data and code are available at the project website\footnote{\url{https://sites.google.com/view/drl-necsa}}. | We propose NECSA, a simple and effective state abstraction-based episodic control containing a more comprehensive episodic memory, a novel state measurement, and a multi-step state analysis. |
In this paper, we work towards a neural surrogate to model wireless electro-magnetic propagation effects in indoor environments.
Such neural surrogates provide a fast, differentiable, and continuous representation of the environment and enables end-to-end optimization for downstream tasks (e.g., network planning). Specifically, the goal of the paper is to render the wireless signal (e.g., time-of-flights, power of each path) in an environment as a function of the sensor's spatial configuration (e.g., placement of transmit and receive antennas). NeRF-based approaches have shown promising results in the visual setting (RGB image signal, with a camera sensor), where the key idea is to algorithmically evaluate the 'global' signal (e.g., using volumetric rendering) by breaking it down in a sequence of 'local' evaluations (e.g., using co-ordinate neural networks). In a similar spirit, we model the time-angle channel impulse response (the global wireless signal) as a superposition of multiple paths. The wireless characteristics (e.g., power) of each path is a result of multiple evaluations of a neural network that learns implicit ray-surface interaction properties. We evaluate our approach in multiple indoor scenarios and demonstrate that our model achieves strong performance (e.g., $<$0.33ns error in time-of-flight predictions). Furthermore, we demonstrate that our neural surrogate whitens the `black-box' wireless simulators, and thus enables inverse rendering applications (e.g., user localization). | Neural wireless ray tracer |
Deep neural networks have been shown vulnerable to adversarial examples. Even though many defence methods have been proposed to enhance the robustness, it is still a long way toward providing an attack-free method to build a trustworthy machine learning system. In this paper, instead of enhancing the robustness, we take the investigator's perspective and propose a new framework to trace the first compromised model in a forensic investigation manner. Specifically, we focus on the following setting: the machine learning service provider provides models for a set of customers. However, one of the customers conducted adversarial attacks to fool the system. Therefore, the investigator's objective is to identify the first compromised model by collecting and analyzing evidence from only available adversarial examples. To make the tracing viable, we design a random mask watermarking mechanism to differentiate adversarial examples from different models. First, we propose a tracing approach in the data-limited case where the original example is also available. Then, we design a data-free approach to identify the adversary without accessing the original example. Finally, the effectiveness of our proposed framework is evaluated by extensive experiments with different model architectures, adversarial attacks, and datasets. | This paper proposes a forensic mechanism for the aftermath of adversarial attacks. |
This paper studies federated learning (FL)—especially cross-silo FL—with data from people who do not trust the server or other silos. In this setting, each silo (e.g. hospital) has data from different people (e.g. patients) and must maintain the privacy of each person’s data (e.g. medical record), even if the server or other silos act as adversarial eavesdroppers. This requirement motivates the study of Inter-Silo Record-Level Differential Privacy (ISRL-DP), which requires silo $i$’s communications to satisfy record/item-level differential privacy (DP). ISRL-DP ensures that the data of each person (e.g. patient) in silo $i$ (e.g. hospital $i$) cannot be leaked. ISRL-DP is different from well-studied privacy notions. Central and user-level DP assume that people trust the server/other silos. On the other end of the spectrum, local DP assumes that people do not trust anyone at all (even their own silo). Sitting between central and local DP, ISRL-DP makes the realistic assumption (in cross-silo FL) that people trust their own silo, but not the server or other silos. In this work, we provide tight (up to logarithms) upper and lower bounds for ISRL-DP FL with convex/strongly convex loss functions and homogeneous (i.i.d.) silo data. Remarkably, we show that similar bounds are attainable for smooth losses with arbitrary heterogeneous silo data distributions, via an accelerated ISRL-DP algorithm. We also provide tight upper and lower bounds for ISRL-DP federated empirical risk minimization, and use acceleration to attain the optimal bounds in fewer rounds of communication than the state-of-the-art. Finally, with a secure “shuffler” to anonymize silo messages (but without a trusted server), our algorithm attains the optimal central DP rates under more practical trust assumptions. Numerical experiments show favorable privacy-accuracy tradeoffs for our algorithm in classification and regression tasks. | Optimal algorithms for differentially private convex/strongly convex federated learning with data from people who do not trust the server or other silos/clients. |
Causal discovery from time-series data has been a central task in machine learning. Recently, Granger causality inference is gaining momentum due to its good explainability and high compatibility with emerging deep neural networks. However, most existing methods assume structured input data and degenerate greatly when encountering data with randomly missing entries or non-uniform sampling frequencies, which hampers their applications in real scenarios. To address this issue, here we present CUTS, a neural Granger causal discovery algorithm to jointly impute unobserved data points and build causal graphs, via plugging in two mutually boosting modules in an iterative framework: (i) Latent data prediction stage: designs a Delayed Supervision Graph Neural Network (DSGNN) to hallucinate and register unstructured data which might be of high dimension and with complex distribution; (ii) Causal graph fitting stage: builds a causal adjacency matrix with imputed data under sparse penalty. Experiments show that CUTS effectively infers causal graphs from irregular time-series data, with significantly superior performance to existing methods. Our approach constitutes a promising step towards applying causal discovery to real applications with non-ideal observations. | Discovering causal relations from unstructured time-series data with a mutually boosting iterative method. |
Probabilistic logical rule learning has shown great strength in logical rule mining and knowledge graph completion. It learns logical rules to predict missing edges by reasoning on existing edges in the knowledge graph. However, previous efforts have largely been limited to only modeling chain-like Horn clauses such as R1(x; z) ^ R2(z; y) ) H(x; y). This formulation overlooks additional contextual information from neighboring sub-graphs of entity variables x, y and z. Intuitively, there is a large gap here, as local sub-graphs have been found to provide important information for knowledge graph completion. Inspired by these observations, we propose Logical Entity RePresentation (LERP) to encode contextual information of entities in the knowledge graph. A LERP is designed as a vector of probabilistic logical functions on the entity’s neighboring sub-graph. It is an interpretable representation while allowing for differentiable optimization. We can then incorporate LERP into probabilistic logical rule learning to learn more expressive rules. Empirical results demonstrate that with LERP, our model outperforms other rule learning methods in knowledge graph completion and is comparable or even superior to state-of-the-art black-box methods. Moreover, we find that our model can discover a more expressive family of logical rules. LERP can also be further combined with embedding learning methods like TransE to make it more interpretable. | We propose logical entity representation (LERP) to incorporate contextual information of entities into logical rule learning. |
In silico prediction of the ligand binding pose to a given protein target is a crucial but challenging task in drug discovery.
This work focuses on blind flexible self-docking, where we aim to predict the positions, orientations and conformations of docked molecules. Traditional physics-based methods usually suffer from inaccurate scoring functions and high inference cost. Recently, data-driven methods based on deep learning techniques are attracting growing interest thanks to their efficiency during inference and promising performance. These methods usually either adopt a two-stage approach by first predicting the distances between proteins and ligands and then generating the final coordinates based on the predicted distances, or directly predicting the global roto-translation of ligands. In this paper, we take a different route. Inspired by the resounding success of AlphaFold2 for protein structure prediction, we propose E3Bind, an end-to-end equivariant network that iteratively updates the ligand pose. E3Bind models the protein-ligand interaction through careful consideration of the geometric constraints in docking and the local context of the binding site. Experiments on standard benchmark datasets demonstrate the superior performance of our end-to-end trainable model compared to traditional and recently-proposed deep learning methods. | An end-to-end equivariant framework for protein-ligand docking through iterative coordinate refinement with careful consideration of the geometric constraints in docking and the local context of the binding site. |
Modern studies in radiograph representation learning (R$^2$L) rely on either self-supervision to encode invariant semantics or associated radiology reports to incorporate medical expertise, while the complementarity between them is barely noticed. To explore this, we formulate the self- and report-completion as two complementary objectives and present a unified framework based on masked record modeling (MRM). In practice, MRM reconstructs masked image patches and masked report tokens following a multi-task scheme to learn knowledge-enhanced semantic representations. With MRM pre-training, we obtain pre-trained models that can be well transferred to various radiography tasks. Specifically, we find that MRM offers superior performance in label-efficient fine-tuning. For instance, MRM achieves 88.5% mean AUC on CheXpert using 1% labeled data, outperforming previous R$^2$L methods with 100% labels. On NIH ChestX-ray, MRM outperforms the best performing counterpart by about 3% under small labeling ratios. Besides, MRM surpasses self- and report-supervised pre-training in identifying the pneumonia type and the pneumothorax area, sometimes by large margins. | We propose to learn radiograph representations via masked record modeling. |
We introduce a simple modification to the standard maximum likelihood estimation (MLE) framework. Rather than maximizing a single unconditional likelihood of the data under the model, we maximize a family of \textit{noise conditional} likelihoods consisting of the data perturbed by a continuum of noise levels. We find that models trained this way are more robust to noise, obtain higher test likelihoods, and generate higher quality images. They can also be sampled from via a novel score-based sampling scheme which combats the classical \textit{covariate shift} problem that occurs during sample generation in autoregressive models. Applying this augmentation to autoregressive image models, we obtain 3.32 bits per dimension on the ImageNet 64x64 dataset, and substantially improve the quality of generated samples in terms of the Frechet Inception distance (FID) --- from 37.50 to 12.09 on the CIFAR-10 dataset. | We propose a noise-robust modification for maximum likelihood estimation. Under this framework, we improve density estimation and significantly enhance the sample quality of images generated by autoregressive models. |
Most robust aggregators for distributed or federated learning assume that adversarial clients are the minority in the system. In contrast, this paper considers the majority adversary setting. We first show that a filtering method using a few trusted clients can defend against many standard attacks. However, a new attack called Mimic-Shift can circumvent simple filtering. To this end, we develop a re-weighting strategy that identifies and down-weights the potential adversaries under the majority adversary regime. We show that our aggregator converges to a neighborhood around the optimum under the Mimic-Shift attack. Empirical results further show that our aggregator achieves negligible accuracy loss with a majority of adversarial clients, outperforming strong baselines. | This paper shows two methods for improving the adversarial robustness of federated learning under a majority adversary regime. |
The development of CLIP [Radford et al., 2021] has sparked a debate on whether adding language supervision can yield vision models with more transferable representations than traditional image-only methods. Our work studies this question through a carefully controlled comparison of two approaches, in terms of their ability to learn representations that generalize to downstream classification tasks. We find that when the pre-training data meets certain criteria---it is sufficiently large and contains descriptive captions with low variability----image-only methods do not match CLIP's performance even when they are trained with more image data. However, contrary to what one might expect, there are practical settings in which these criteria are not met, wherein added supervision through captions is actually detrimental.
Motivated by our findings, we devise simple data and algorithmic interventions to improve the transfer performance of CLIP-style models. | Our work performs a systematic investigation into whether additional language supervision (in CLIP) helps models learn more transferrable representations. |
Training deep neural networks for classification often includes minimizing the training loss beyond the zero training error point. In this phase of training, a “neural collapse” behavior has been observed: the variability of features (outputs of the penultimate layer) of within-class samples decreases and the mean features of different classes approach a certain tight frame structure. Recent works analyze this behavior via idealized unconstrained features models where all the minimizers exhibit exact collapse. However, with practical networks and datasets, the features typically do not reach exact collapse, e.g., because deep layers cannot arbitrarily modify intermediate features that are far from being collapsed. In this paper, we propose a richer model that can capture this phenomenon by forcing the features to stay in the vicinity of a predefined features matrix (e.g., intermediate features). We explore the model in the small vicinity case via perturbation analysis and establish results that cannot be obtained by the previously studied models. For example, we prove reduction in the within-class variability of the optimized features compared to the predefined input features (via analyzing gradient flow on the “central-path” with minimal assumptions), analyze the minimizers in the near-collapse regime, and provide insights on the effect of regularization hyperparameters on the closeness to collapse. We support our theory with experiments in practical deep learning settings. | We propose a new model for exploring practical NC behavior and establish, via perturbation analysis, results that could not have been obtained by existing (idealized) models. |
Molecular representation learning plays a crucial role in AI-assisted drug discovery research. Encoding 3D molecular structures through Euclidean neural networks has become the prevailing method in the geometric deep learning community. However, the equivariance constraints and message passing in Euclidean space may limit the network expressive power. In this work, we propose a Harmonic Molecular Representation learning (HMR) framework, which represents a molecule using the Laplace-Beltrami eigenfunctions of the molecular surface. HMR offers a multi-resolution representation of molecular geometric and chemical properties on 2D Riemannian manifold. We also introduce a harmonic message passing method to realize efficient spectral message passing over the surface manifold for better molecular encoding. Our proposed method shows comparable predictive power to current models in small molecule property prediction, and outperforms the state-of-the-art deep learning models for the rigid protein docking challenge, demonstrating its versatility in molecular representation learning. | We propose a harmonic molecular representation learning framework to achieve multi-resolution molecular encoding on 2D Riemannian manifold. |
As an intrinsic and fundamental property of big data, data heterogeneity exists in a variety of real-world applications, such as in agriculture, sociology, health care, etc. For machine learning algorithms, the ignorance of data heterogeneity will significantly hurt the generalization performance and the algorithmic fairness, since the prediction mechanisms among different sub-populations are likely to differ. In this work, we focus on the data heterogeneity that affects the prediction of machine learning models, and first formalize the Predictive Heterogeneity, which takes into account the model capacity and computational constraints. We prove that it can be reliably estimated from finite data with PAC bounds even in high dimensions. Additionally, we propose the Information Maximization (IM) algorithm, a bi-level optimization algorithm, to explore the predictive heterogeneity of data. Empirically, the explored predictive heterogeneity provides insights for sub-population divisions in agriculture, sociology, and object recognition, and leveraging such heterogeneity benefits the out-of-distribution generalization performance. | In this work, we propose the predictive heterogeneity to measure the data heterogeneity that affects prediction. Theoretical analysis and empirical results validate the rationality of the proposed measure. |
Autoencoding has been a popular and effective framework for learning generative models for images, with much empirical success. Autoencoders often use generic deep networks as the encoder and decoder, which are difficult to interpret, and the learned representations lack clear structure. In this work, we replace the encoder and decoder with standard convolutional sparse coding and decoding layers, obtained from unrolling an optimization algorithm for solving a (convexified) sparse coding program. Furthermore, to avoid computational difficulties in minimizing distributional distance between the real and generated images, we utilize the recent closed-loop transcription (CTRL) framework that maximizes the rate reduction of the learned sparse representations. We show that such a simple framework demonstrates surprisingly competitive performance on large datasets, such as ImageNet-1K, compared to existing autoencoding and generative methods under fair conditions. Even with simpler networks and less computational resources, our method demonstrates splendid visual quality in regenerated images with striking sample-wise consistency. More surprisingly, the learned autoencoder generalizes to unseen datasets. Our method enjoys several side benefits, including more structured and interpretable representations, more stable convergence, scalability to large datasets -- indeed, our method is the first sparse coding generative method to scale up to ImageNet -- and trainability with smaller batch sizes. | This paper combines the recent closed-loop transcription framework with convolutional sparse coding layers and demonstrates superior generative autoencoding performance. |
The ability to computationally generate novel yet physically foldable protein structures could lead to new biological discoveries and new treatments targeting yet incurable diseases. Despite recent advances in protein structure prediction, directly generating diverse, novel protein structures from neural networks remains difficult. In this work, we present a new diffusion-based generative model that designs protein backbone structures via a procedure that mirrors the native folding process. We describe protein backbone structure as a series of consecutive angles capturing the relative orientation of the constituent amino acid residues, and generate new structures by denoising from a random, unfolded state towards a stable folded structure. Not only does this mirror how proteins biologically twist into energetically favorable conformations, the inherent shift and rotational invariance of this representation crucially alleviates the need for complex equivariant networks. We train a denoising diffusion probabilistic model with a simple transformer backbone and demonstrate that our resulting model unconditionally generates highly realistic protein structures with complexity and structural patterns akin to those of naturally-occurring proteins. As a useful resource, we release the first open-source codebase and trained models for protein structure diffusion. | Inspired by the protein folding process, we introduce a new diffusion-based generative model that acts on the inter-residue angles in protein backbones and generates diverse, designable protein structures without needing equivariance mechanisms. |
The computation of Wasserstein gradient direction is essential for posterior sampling problems and scientific computing. The approximation of the Wasserstein gradient with finite samples requires solving a variational problem. We study the variational problem in the family of two-layer networks with squared-ReLU activations, towards which we derive a semi-definite programming (SDP) relaxation. This SDP can be viewed as an approximation of the Wasserstein gradient in a broader function family including two-layer networks. By solving the convex SDP, we obtain the optimal approximation of the Wasserstein gradient direction in this class of functions. Numerical experiments including PDE-constrained Bayesian inference and parameter estimation in COVID-19 modeling demonstrate the effectiveness of the proposed method. | Wasserstein gradient descent meets neural networks and convex optimization |
Recent work has observed an intriguing "Neural Collapse" phenomenon in well-trained neural networks, where the last-layer representations of training samples with the same label collapse into each other. This suggests that the last-layer representations are completely determined by the labels, and do not depend on the intrinsic structure of input distribution. We provide evidence that this is not a complete description, and that the apparent collapse hides important fine-grained structure in the representations. Specifically, even when representations apparently collapse, the small amount of remaining variation can still faithfully and accurately captures the intrinsic structure of input distribution. As an example, if we train on CIFAR-10 using only 5 coarse-grained labels (by combining two classes into one super-class) until convergence, we can reconstruct the original 10-class labels from the learned representations via unsupervised clustering. The reconstructed labels achieve $93\%$ accuracy on the CIFAR-10 test set, nearly matching the normal CIFAR-10 accuracy for the same architecture. Our findings show concretely how the structure of input data can play a significant role in determining the fine-grained structure of neural representations, going beyond what Neural Collapse predicts.
| We provide compelling empirical evidence proving that there exists fine-grained structures in the last-layer representations of a well trained neural network, as a complement to existing Neural Collapse hypothesis. |
Winning tickets are sparse subnetworks of a deep network that can be trained in isolation to the same performance as the full network. Winning tickets have been found in many different contexts, however their structural characteristics are not well understood. We propose that the signs of the connections in winning tickets play a crucial role. We back this claim by introducing a sign-based structural comparison
metric that allows to distinguish winning tickets from other sparse networks. We further analyze typical (signed) patterns in convolutional kernels of winning tickets and find structures that resemble patterns found in trained networks. | Winning tickets show structural similarities when taking signs of connections into account. |
Thanks to the large pre-trained vision-language models (VLMs) like CLIP, we can craft a zero-shot classifier by ``prompt'', e.g., using the model provided similarity measure between an image and the prompt sentence ``$\texttt{a photo of a [CLASS]}$'', as the confidence score of predicting the image is ``$\texttt{[CLASS]}$''. Therefore, prompt shows a great potential for fast adapting the VLMs to downstream tasks if we fine-tune the prompt-based similarity measure. However, we find a common failure that improper fine-tuning may not only undermine the prompt's inherent prediction for the task-related classes, but also for other classes in the VLM vocabulary. Existing methods still address this problem by using traditional anti-overfitting techniques such as early stopping and data augmentation, which lack a principled solution specific to prompt. We present Prompt-aligned Gradient, dubbed $\texttt{ProGrad}$, to prevent prompt tuning from forgetting the the general knowledge learned from VLMs. In particular, $\texttt{ProGrad}$ only updates the prompt whose gradient is aligned (or non-conflicting) to the ``general direction'', which is represented as the gradient of the KL loss of the pre-defined prompt prediction. Extensive experiments demonstrate the stronger few-shot generalization ability of $\texttt{ProGrad}$ over state-of-the-art prompt tuning methods. Codes are in Appendix. | We present Prompt-aligned Gradient to prevent prompt tuning from forgetting the general knowledge learned from CLIP. |
Different approaches to eXplainable Artificial Intelligence (XAI) have been explored including (1) the systematic study of the effect of individual training data sample on the final model (2) posthoc attribution methods that assign importance values to the components of each data sample. Combining concepts from both approaches, we introduce kaBEDONN, a system of ordered dataset coupled with a posthoc and model-agnostic method for querying \textit{relevant} training data samples. These \textit{relevant} data are intended as the explanations for model predictions that are both user-friendly and easily adjustable by developers. Explanations can thus be finetuned and damage control can be performed with ease. | A posthoc method for providing explanation to blackbox algorithms by querying similar data |
Universal Adversarial Perturbations (UAPs) are imperceptible, image-agnostic vectors that cause deep neural networks (DNNs) to misclassify inputs from a data distribution with high probability. In practical attack scenarios, adversarial perturbations may undergo transformations such as changes in pixel intensity, rotation, etc. while being added to DNN inputs. Existing methods do not create UAPs robust to these real-world transformations, thereby limiting their applicability in attack scenarios. In this work, we introduce and formulate robust UAPs. We build an iterative algorithm using probabilistic robustness bounds and transformations generated by composing arbitrary sub-differentiable transformation functions to construct such robust UAPs. We perform an extensive evaluation on the popular CIFAR-10 and ILSVRC 2012 datasets measuring our UAPs' robustness under a wide range common, real-world transformations such as rotation, contrast changes, etc. Our results show that our method can generate UAPs up to 23% more robust than existing state-of-the-art baselines. | This paper introduces the concept of Robust Universal Adversarial Perturbations and a new algorithm, RobustUAP, which can be used to generate UAPs robust under human-interpretable, real-word transformations, such as rotation, contrast changes, etc. |
Offline reinforcement learning (RL) promises the ability to learn effective policies solely using existing, static datasets, without any costly online interaction. To do so, offline RL methods must handle distributional shift between the dataset and the learned policy. The most common approach is to learn conservative, or lower-bound, value functions, which underestimate the return of OOD actions. However, such methods exhibit one notable drawback: policies optimized on such value functions can only behave according to a fixed, possibly suboptimal, degree of conservatism. However, this can be alleviated if we instead are able to learn policies for varying degrees of conservatism at training time and devise a method to dynamically choose one of them during evaluation. To do so, in this work, we propose learning value functions that additionally condition on the degree of conservatism, which we dub confidence-conditioned value functions. We derive a new form of a Bellman backup that simultaneously learns Q-values for any degree of confidence with high probability. By conditioning on confidence, our value functions enable adaptive strategies during online evaluation by controlling for confidence level using the history of observations thus far. This approach can be implemented in practice by conditioning the Q-function from existing conservative algorithms on the confidence. We theoretically show that our learned value functions produce conservative estimates of the true value at any desired confidence. Finally, we empirically show that our algorithm outperforms existing conservative offline RL algorithms on multiple discrete control domains. | We propose a new offline reinforcement learning algorithm that adapts how conservative its behavior will be. |
Differentiable neural architecture search (DARTS) is a popular method for neural architecture search (NAS), which performs cell-search and utilizes continuous relaxation to improve the search efficiency via gradient-based optimization. The main shortcoming of DARTS is performance collapse, where the discovered architecture suffers from a pattern of declining quality during search. Performance collapse has become an important topic of research, with many methods trying to solve the issue through either regularization or fundamental changes to DARTS.
However, the weight-sharing framework used for cell-search in DARTS and the convergence of architecture parameters has not been analyzed yet. In this paper, we provide a thorough and novel theoretical and empirical analysis on DARTS and its point of convergence.
We show that DARTS suffers from a specific structural flaw due to its weight-sharing framework that limits the convergence of DARTS to saturation points of the softmax function. This point of convergence gives an unfair advantage to layers closer to the output in choosing the optimal architecture, causing performance collapse. We then propose two new regularization terms that aim to prevent performance collapse by harmonizing operation selection via aligning gradients of layers.
Experimental results on six different search spaces and three different datasets show that our method ($\Lambda$-DARTS) does indeed prevent performance collapse, providing justification for our theoretical analysis and the proposed remedy. | We pinpoint the reason for performance collapse in DARTS and provide theoretical and empirical analysis on that as well as a solution to remedy the performance collapse via harmonizing the decisions of different cell. |
Recently, great success has been made in learning visual representations from text supervision, facilitating the emergence of text-supervised semantic segmentation. However, existing works focus on pixel grouping and cross-modal semantic alignment, while ignoring the correspondence among multiple augmented views of the same image. To overcome such limitation, we propose multi-View Consistent learning (ViewCo) for text-supervised semantic segmentation. Specifically, we first propose text-to-views consistency modeling to learn correspondence for multiple views of the same input image. Additionally, we propose cross-view segmentation consistency modeling to address the ambiguity issue of text supervision by contrasting the segment features of Siamese visual encoders. The text-to-views consistency benefits dense assignment of the visual features by encouraging different crops to align with the same text, while the cross-view segmentation consistency modeling provides additional self-supervision, overcoming the limitation of ambiguous text supervision for segmentation masks. Trained with large-scale image-text data, our model can directly segment objects of arbitrary categories in a zero-shot manner. Extensive experiments show that ViewCo outperforms state-of-the-art methods on average by up to 2.9%, 1.6%, and 2.4% mIoU on PASCAL VOC2012, PASCAL Context, and COCO, respectively. | Discovering text-supervised segmentation masks via multi-view semantic consistency |
This paper proposes a generative model that implements cooperative adversarial learning via closed-loop transcription. In the generative model training, the encoder and decoder are trained simultaneously, and not only the adversarial process but also a cooperative process is included. In the adversarial process, the encoder plays as a critic to maximize the distance between the original and transcribed images, in which the distance is measured by rate reduction in the feature space; in the cooperative process, the encoder and the decoder cooperatively minimize the distance to improve the transcription quality. Cooperative adversarial learning possesses the concepts and properties of Auto-Encoding and GAN, and it is unique in that the encoder actively controls the training process as it is trained in both learning processes in two different roles. Experiments demonstrate that without regularization techniques, our generative model is robust to net architectures and easy to train, sample-wise reconstruction performs well in terms of sample features, and disentangled visual attributes are well modeled in independent principal components. | This paper proposes a generative model that implements cooperative adversarial learning, which is robust to net architectures and performs well, and disentangled visual attributes are well modeled in independent principal components. |
We present TabPFN, a trained Transformer that can do supervised classification for small tabular datasets in less than a second, needs no hyperparameter tuning and is competitive with state-of-the-art classification methods.
TabPFN is fully entailed in the weights of our network, which accepts training and test samples as a set-valued input and yields predictions for the entire test set in a single forward pass.
TabPFN is a Prior-Data Fitted Network (PFN) and is trained offline once, to approximate Bayesian inference on synthetic datasets drawn from our prior.
This prior incorporates ideas from causal reasoning: It entails a large space of structural causal models with a preference for simple structures.
On the $18$ datasets in the OpenML-CC18 suite that contain up to 1000 training data points, up to 100 purely numerical features without missing values, and up to 10 classes, we show that our method clearly outperforms boosted trees and performs on par with complex state-of-the-art AutoML systems with up to $230\times$ speedup.
This increases to a $5\,700\times$ speedup when using a GPU. We also validate these results on an additional 67 small numerical datasets from OpenML.
We provide all our code, the trained TabPFN, an interactive browser demo and a Colab notebook at https://github.com/automl/TabPFN. | We present TabPFN, a trained Transformer that learned to solve small tabular data classification problems at SOTA level in less than a second by training on synthetic data generated by integrating principles from causal reasoning and simplicity. |
We present F-VLM, a simple open-vocabulary object detection method built uponFrozenVision andLanguageModels. F-VLM simplifies the current multi-stage training pipeline by eliminating the need for knowledge distillation or detection-tailored pretraining. Surprisingly, we observe that a frozen VLM: 1) retains the locality-sensitive features necessary for detection, and 2) is a strong region classifier. We finetune only the detector head and combine the detector and VLM outputs for each region at inference time. F-VLM shows compelling scaling behavior and achieves +6.5 mask AP improvement over the previous state of theart on novel categories of LVIS open-vocabulary detection benchmark. In addition, we demonstrate very competitive results on COCO open-vocabulary detection benchmark and cross-dataset transfer detection, in addition to significant training speed-up and compute savings. Code will be released.
| We propose a novel open-vocabulary detection approach by building upon frozen vision and language models. |
Hierarchical reinforcement learning is a promising approach that uses temporal abstraction to solve complex long horizon problems. However, simultaneously learning a hierarchy of policies is unstable as it is challenging to train higher-level policy when the lower-level primitive is non-stationary. In this paper, we propose to generate a curriculum of achievable subgoals for evolving lower-level primitives using reinforcement learning and imitation learning. The lower level primitive periodically performs data relabeling on a handful of expert demonstrations using our primitive informed parsing method. We derive expressions to bound the sub-optimality of our method and develop a practical algorithm for hierarchical reinforcement learning. Since our approach uses a handful of expert demonstrations, it is suitable for most real world robotic control tasks. Experimental results on complex maze navigation and robotic manipulation environments show that inducing hierarchical curriculum learning significantly improves sample efficiency, and results in better learning of goal conditioned policies in complex temporally extended tasks. | We effectively leverage expert demonstrations using our curriculum learning based approach to deal with non-stationarity in the context of hierarchical reinforcement learning. |
Content creators compete for user attention. Their reach crucially depends on algorithmic choices made by developers on online platforms. To maximize exposure, many creators adapt strategically, as evidenced by examples like the sprawling search engine optimization industry. This begets competition for the finite user attention pool. We formalize these dynamics in what we call an exposure game, a model of incentives induced by modern algorithms including factorization and (deep) two-tower architectures. We prove that seemingly innocuous algorithmic choices—e.g., non-negative vs. unconstrained factorization—significantly affect the existence and character of (Nash) equilibria in exposure games. We proffer use of creator behavior models like ours for an (ex-ante) pre-deployment audit. Such an audit can identify misalignment between desirable and incentivized content, and thus complement post-hoc measures like content filtering and moderation. To this end, we propose tools for numerically finding equilibria in exposure games, and illustrate results of an audit on the MovieLens and LastFM datasets. Among else, we find that the strategically produced content exhibits strong dependence between algorithmic exploration and content diversity, and between model expressivity and bias towards gender-based user and creator groups. | Algorithmic choices in modern recommenders may have significant and unexpected effects on content creator incentives. |
Neural networks are known to be highly sensitive to adversarial examples. These may arise due to different factors, such as random initialization, or spurious correlations in the learning problem. To better understand these factors, we provide a precise study of the adversarial robustness in different scenarios, from initialization to the end of training in different regimes, as well as intermediate scenarios where initialization still plays a role due to “lazy” training. We consider over-parameterized networks in high dimensions with quadratic targets and infinite samples. Our analysis allows us to identify new tradeoffs between approximation (as measured via test error) and robustness, whereby robustness can only get worse when test error improves, and vice versa. We also show how linearized lazy training
regimes can worsen robustness, due to improperly scaled random initialization. Our theoretical results are illustrated with numerical experiments. | Tradeoffs between test-error and robustness for 2-layer neural networks in different learning regimes (RF, lazy training, SGD) |
Modern classification neural networks are notoriously prone to being overly confident in their predictions. With multiple calibration methods having been proposed so far, there has been noteworthy progress in reducing this overconfidence. However, to the best of our knowledge, prior methods have exclusively focused on the factors that affect calibration, leaving open the reverse question of how (mis)calibration impacts network training. Aiming for a better understanding of this interplay, we propose a temperature-based Cooling method for calibrating classification neural networks during training. Cooling has a substantial effect on the gradients and reduces the need for a learning rate schedule. We investigate different variants of Cooling, with the simplest one, last layer Cooling, being also the best-performant one, improving network performance on a range of datasets, network architectures, and hyperparameter settings. | The paper proposes a calibration method applied during neural network training, which removes the need for a learning rate schedule. |
Conditional language models are predominantly trained with maximum likelihood estimation (MLE), giving probability mass to sparsely observed target sequences. While MLE trained models assign high probability to plausible sequences given the context, the model probabilities often do not accurately rank-order generated sequences by quality. This has been empirically observed in beam search decoding as output quality degrading with large beam sizes, and decoding strategies benefiting from heuristics such as length normalization and repetition-blocking. In this work, we introduce sequence likelihood calibration (SLiC) where the likelihood of model generated sequences are calibrated to better align with reference sequences in the model’s latent space. With SLiC, decoding heuristics become unnecessary and decoding candidates’ quality significantly improves regardless of the decoding method. Furthermore, SLiC shows no sign of diminishing returns with model scale, and presents alternative ways to improve quality with limited training and inference budgets. With SLiC, we exceed or match SOTA results on a wide range of generation tasks spanning abstractive summarization, question generation, abstractive question answering and data-to-text generation, even with modest-sized models. | A proposed sequence likelihood calibration stage improves fine-tuned conditional language models, leading to new state-of-the-art results in abstractive summarization, question generation, abstractive question answering and data-to-text. |
Generative flow networks (GFlowNets) are a family of algorithms for training a sequential sampler of discrete objects under an unnormalized target density and have been successfully used for various probabilistic modeling tasks. Existing training objectives for GFlowNets are either local to states or transitions, or propagate a reward signal over an entire sampling trajectory. We argue that these alternatives represent opposite ends of a gradient bias-variance tradeoff and propose a way to exploit this tradeoff to mitigate its harmful effects. Inspired by the TD($\lambda$) algorithm in reinforcement learning, we introduce subtrajectory balance or SubTB($\lambda$), a GFlowNet training objective that can learn from partial action subsequences of varying lengths. We show that SubTB($\lambda$) accelerates sampler convergence in previously studied and new environments and enables training GFlowNets in environments with longer action sequences and sparser reward landscapes than what was possible before. We also perform a comparative analysis of stochastic gradient dynamics, shedding light on the bias-variance tradeoff in GFlowNet training and the advantages of subtrajectory balance. | GFlowNet training is made faster and more stable by learning from subtrajectories. |
Efficient exploration and model-building are critical for learning in large state- spaces. However, agents typically face problems like getting stuck locally during exploration and catastrophic forgetting in their construction of models when the environments are heterogeneous. Here, we propose and apply the concept of Fragmentation-and-Recall to solve spatial (FarMap) and reinforcement learning problems (FarCuriosity). Agents construct local maps or local models, respectively, which are used to predict the current observation. High surprisal points lead to a fragmentation event. At fracture points, we store the current map or model fragment in a long-term memory (LTM) and initialize a new fragment. On the other hand, Fragments are recalled (and thus reused) from LTM if the observations of their fracture points match the agent’s current observation during exploration. The set of fracture points defines a set of intrinsic potential subgoals. Agents choose their next subgoal from the set of near and far potential subgoals in the current fragment or LTM, respectively. Thus, local maps and model fragments guide exploration locally and avoid catastrophic forgetting in learning heterogeneous environments, while LTM promotes exploration more globally. We evaluate FarMap and FarCuriosity on complex procedurally-generated spatial environments and on reinforcement learning benchmarks and demonstrate that the proposed methods are more efficient at exploration and memory use, and in harvesting extrinsic rewards, respectively. | We propose a novel framework for exploration based on fragmentation-and-recall. |
We present the Group Propagation Vision Transformer (GPViT): a novel non- hierarchical (i.e. non-pyramidal) transformer model designed for general visual recognition with high-resolution features. High-resolution features (or tokens) are a natural fit for tasks that involve perceiving fine-grained details such as detection and segmentation, but exchanging global information between these features is expensive in memory and computation because of the way self-attention scales. We provide a highly efficient alternative Group Propagation Block (GP Block) to exchange global information. In each GP Block, features are first grouped to- gether by a fixed number of learnable group tokens; we then perform Group Propagation where global information is exchanged between the grouped fea- tures; finally, global information in the updated grouped features is returned back to the image features through a transformer decoder. We evaluate GPViT on a variety of visual recognition tasks including image classification, semantic seg- mentation, object detection, and instance segmentation. Our method achieves significant performance gains over previous works across all tasks, especially on tasks that require high-resolution outputs, for example, our GPViT-L3 out- performs Swin Transformer-B by 2.0 mIoU on ADE20K semantic segmentation with only half as many parameters. Code and pre-trained models are available at https://github.com/ChenhongyiYang/GPViT. | A high-resolution vision transformer architecture based on a new efficient global information exchange mechanism for general visual recognition. |
Transfer learning is the predominant paradigm for training deep networks on small target datasets. Models are typically pretrained on large “upstream” datasets for classification, as such labels are easy to collect, and then finetuned on downstream” tasks such as action localisation, which are smaller due to their finer-grained annotations.
In this paper, we question this approach, and propose co-finetuning -- simultaneously training a single model on multiple “upstream” and “downstream” tasks. We demonstrate that co-finetuning outperforms traditional transfer learning when using the same total amount of data, and also show how we can easily extend our approach to multiple “upstream” datasets to further improve performance. In particular, co-finetuning significantly improves the performance on rare classes in our downstream task, as it has a regularising effect, and enables the network to learn feature representations that transfer between different datasets. Finally, we observe how co-finetuning with public, video classification datasets, we are able to achieve state-of-the-art results for spatio-temporal action localisation on the challenging AVA and AVA-Kinetics datasets, outperforming recent works which develop intricate models. | Instead of pretraining on "upstream" datasets and then finetuning on "downstream" tasks, we simultaneously train on all datasets, achieving significant performance improvements across all tasks, and particularly on rare classes. |
Inspired by Regularized Lottery Ticket Hypothesis, which states that competitive smooth (non-binary) subnetworks exist within a dense network, we propose a few-shot class-incremental learning method referred to as Soft-SubNetworks (SoftNet). Our objective is to learn a sequence of sessions incrementally, where each session only includes a few training instances per class while preserving the knowledge of the previously learned ones. SoftNet jointly learns the model weights and adaptive non-binary soft masks at a base training session in which each mask consists of the major and minor subnetwork; the former aims to minimize catastrophic forgetting during training, and the latter aims to avoid overfitting to a few samples in each new training session. We provide comprehensive empirical validations demonstrating that our SoftNet effectively tackles the few-shot incremental learning problem by surpassing the performance of state-of-the-art baselines over benchmark datasets. | SoftNet jointly learns the model weights and adaptive non-binary soft masks at a base training session in which each mask consists of the major and minor subnetwork. |
We propose a novel group disentanglement method called the Context-Aware Variational Autoencoder (CxVAE). Our model can learn disentangled representations on datasets with conditional shift. This phenomenon occurs when the conditional distribution of the instance-level latent variable $\mathbf{z}$ given the input observation $\mathbf{x}$ changes from one group to another (i.e. $p_i(\mathbf{z}|\mathbf{x}) \neq p_j(\mathbf{z}|\mathbf{x})$, where $i,j$ are two different groups). We show that existing methods fail to learn disentangled representations under this scenario because they infer the group $\mathbf{u}$ and instance $\mathbf{z}$ variables separately. CxVAE overcomes this limitation by conditioning the instance inference on the group variable $q(\mathbf{z}|\mathbf{x},\mathbf{u})$. Our model has the novel ability to disentangle ambiguous observations (those with incomplete information about the generative factors), which we evaluate on the task of fair comparisons between student test scores. Additionally, we demonstrate empirically that conditional shift is the cause of our model's improved performance. | A VAE-based model that can group-disentangle data under conditional shift, evaluated on fair comparisons between student test scores. |
Distilling knowledge from a large teacher model to a lightweight one is a widely successful approach for generating compact, powerful models in the semi-supervised learning setting where a limited amount of labeled data is available. In large-scale applications, however, the teacher tends to provide a large number of incorrect soft-labels that impairs student performance. The sheer size of the teacher additionally constrains the number of soft-labels that can be queried due to prohibitive computational and/or financial costs. The difficulty in achieving simultaneous \emph{efficiency} (i.e., minimizing soft-label queries) and \emph{robustness} (i.e., avoiding student inaccuracies due to incorrect labels) hurts the widespread application of knowledge distillation to many modern tasks. In this paper, we present a parameter-free approach with provable guarantees to query the soft-labels of points that are simultaneously informative and correctly labeled by the teacher. At the core of our work lies a game-theoretic formulation that explicitly considers the inherent trade-off between the informativeness and correctness of input instances. We establish bounds on the expected performance of our approach that hold even in worst-case distillation instances. We present empirical evaluations on popular benchmarks that demonstrate the improved distillation performance enabled by our work relative to that of state-of-the-art active learning and active distillation methods. | A new way of actively soft-labeling points in semi-supervised knowledge distillation to teach the student model in an efficient and robust way |
We propose the Deep Graph Neural Diffusion (DeepGRAND), a class of continuous-depth graph neural networks based on the diffusion process on graphs. DeepGRAND leverages a data-dependent scaling term and a perturbation to the graph diffusivity to make the real part of all eigenvalues of the diffusivity matrix become negative, which ensures two favorable theoretical properties: (i) the node representation does not exponentially converge to a constant vector as the model depth increases, thus alleviating the over-smoothing issue; (ii) the stability of the model is guaranteed by controlling the norm of the node representation. Compared to the baseline GRAND, DeepGRAND mitigates the accuracy drop-off with increasing depth and improves the overall accuracy of the model. We empirically corroborate the advantage of DeepGRAND over many existing graph neural networks on various graph deep learning benchmark tasks. | We propose the Deep Graph Neural Diffusion (DeepGRAND), a class of continuous-depth graph neural networks that leverages a data-dependent scaling term and a perturbation to the graph diffusivity to overcome the oversmoothing issue. |
Conventional representation learning algorithms for knowledge graphs (KG) map each entity to a unique embedding vector, ignoring the rich information contained in the neighborhood. We propose a method named StarGraph, which gives a novel way to utilize the neighborhood information for large-scale knowledge graphs to obtain entity representations. An incomplete two-hop neighborhood subgraph for each target node is at first generated, then processed by a modified self-attention network to obtain the entity representation, which is used to replace the entity embedding in conventional methods. We achieved SOTA performance on ogbl-wikikg2 and got competitive results on fb15k-237. The experimental results proves that StarGraph is efficient in parameters, and the improvement made on ogbl-wikikg2 demonstrates its great effectiveness of representation learning on large-scale knowledge graphs. | knowledge representation learning based on incomplete local subgraph |
There is a rising interest in further exploring the zero-shot learning potential of large pre-trained language models (PLMs). A new paradigm called data-generation-based zero-shot learning has achieved impressive success. In this paradigm, the synthesized data from the PLM acts as the carrier of knowledge, which is used to train a task-specific model with orders of magnitude fewer parameters than the PLM, achieving both higher performance and efficiency than prompt-based zero-shot learning methods on PLMs. The main hurdle of this approach is that the synthesized data from PLM usually contains a significant portion of low-quality samples. Fitting on such data will greatly hamper the performance of the task-specific model, making it unreliable for deployment. Previous methods remedy this issue mainly by filtering synthetic data using heuristic metrics(e.g., output confidence), or refining the data with the help of a human expert, which comes with excessive manual tuning or expensive costs. In this paper, we propose a novel noise-robust re-weighting framework SunGen to automatically construct high-quality data for zero-shot classification problems. Our framework features the ability to learn the sample weights indicating data quality without requiring any human annotation. We theoretically and empirically verify the ability of our method to help construct good-quality synthetic datasets. Notably, SunGen-LSTM yields a 9.8% relative improvement than the baseline on average accuracy across eight different established text classification tasks. | This paper proposes a framework to automatically enhance the quality of PLM-generated data for efficient zero-shot learning, without relying on any human annotation. |
We introduce CAN, a simple, efficient and scalable method for self-supervised learning of visual representations. Our framework is a minimal and conceptually clean synthesis of (C) contrastive learning, (A) masked autoencoders, and (N) the noise prediction approach used in diffusion models. The learning mechanisms are \emph{complementary} to one another: contrastive learning shapes the embedding space across a batch of image samples; masked autoencoders focus on reconstruction of the low-frequency spatial correlations in a single image sample; and noise prediction encourages the reconstruction of the high-frequency components of an image. The combined approach results in a robust, scalable and simple-to-implement algorithm. The training process is symmetric, with $50\%$ of patches in \emph{both views} being masked at random, yielding a considerable efficiency improvement over prior contrastive learning methods. Extensive empirical studies on linear evaluation, finetuning, transfer learning, and robustness demonstrate that our approach achieves strong downstream performance. For instance, when pre-training ViT-B encoders on the curated ImageNet dataset, CAN achieves $74.8\%$ top-1 linear probing accuracy, an absolute improvement of $6.8\%$ over MAE and $1.3\%$ over SimCLR with the same architecture and data augmentations. CAN is especially useful for pre-training on larger uncurated datasets such as JFT-300M: the finetuned performance on ImageNet of our ViT-L model is $85.9\%$, compared to $85.0\%$ for SimCLR, and $85.4\%$ for MAE. For linear probe on ImageNet, CAN achieves $75.4\%$ compared to $71.8\%$ for SimCLR and $64.1\%$ for MAE. The overall FLOPs load is $41\%$ \emph{lower} than SimCLR\footnote{Our code will be released at \url{www.xxx.yyy}.}. | We propose a minimal and conceptually clean synthesis of (C) contrastive learning, (A) masked autoencoders, and (N) the noise prediction for self-supervised learning on images |
This paper introduces ``Succinct Compression”, a method to provide lossless compression of Deep Neural Network (DNN) models for fast and memory-efficient inference. The key insight of our method leverages the concept of \textit{Succinct Data Structures}, which supports fast queries without decompressing the compressed representations. Our method consists of three new insights. First, we introduce two basic building blocks to formulate DNN models, and how they can be extended to be synergistic with compressed models (e.g. pruned or quantized models). Then, we propose a scheme to enable mixed-formulation inference for different layers, to better extract its benefits. Finally, our method exploits a specialized execution pipeline to incorporate different model formulations for fast inference. We quantitatively demonstrate that: our method can (1) enable faster and more memory-efficient inference on uncompressed models; (2) be synergistic with a variety of structure-altered/unaltered compression schemes with better speedup and compression ratio, while preserving the accuracy; and (3) can outperform all other state-of-the-art Model Coding approaches.
| First work to introduce Succinct Data Structures for Fast and Memory-Efficient Computations of Deep Neural Networks |
Finding unified complexity measures and algorithms for sample-efficient learning is a central topic of research in reinforcement learning (RL). The Decision-Estimation Coefficient (DEC) is recently proposed by Foster et al. (2021) as a necessary and sufficient complexity measure for sample-efficient no-regret RL. This paper makes progress towards a unified theory for RL with the DEC framework. First, we propose two new DEC-type complexity measures: Explorative DEC (EDEC), and Reward-Free DEC (RFDEC). We show that they are necessary and sufficient for sample-efficient PAC learning and reward-free learning, thereby extending the original DEC which only captures no-regret learning. Next, we design new unified sample-efficient algorithms for all three learning goals. Our algorithms instantiate variants of the Estimation-To-Decisions (E2D) meta-algorithm with a strong and general model estimation subroutine. Even in the no-regret setting, our algorithm \textsc{E2D-TA} improves upon the algorithms of Foster et al. (2021) which require either bounding a variant of the DEC which may be prohibitively large, or designing problem-specific estimation subroutines. As applications, we recover existing and obtain new sample-efficient learning results for a wide range of tractable RL problems using essentially a single algorithm. Finally, as a connection, we re-analyze two existing optimistic model-based algorithms based on Posterior Sampling or Maximum Likelihood Estimation, showing that they enjoy similar regret bounds as \textsc{E2D-TA} under similar structural conditions as the DEC. | We design new unified algorithms for no-regret, PAC, and reward-free reinforcement learning with general model classes, building on the Decision-Estimation Coefficient and a strong model estimation procedure. |
The message passing-based graph neural networks (GNNs) have achieved great success in many real-world applications.
However, training GNNs on large-scale graphs suffers from the well-known neighbor explosion problem, i.e., the exponentially increasing dependencies of nodes with the number of message passing layers. Subgraph-wise sampling methods---a promising class of mini-batch training techniques---discard messages outside the mini-batches in backward passes to avoid the neighbor explosion problem at the expense of gradient estimation accuracy. This poses significant challenges to their convergence analysis and convergence speeds, which seriously limits their reliable real-world applications. To address this challenge, we propose a novel subgraph-wise sampling method with a convergence guarantee, namely Local Message Compensation (LMC). To the best of our knowledge, LMC is the {\it first} subgraph-wise sampling method with provable convergence. The key idea of LMC is to retrieve the discarded messages in backward passes based on a message passing formulation of backward passes. By efficient and effective compensations for the discarded messages in both forward and backward passes, LMC computes accurate mini-batch gradients and thus accelerates convergence. We further show that LMC converges to first-order stationary points of GNNs. Experiments on large-scale benchmark tasks demonstrate that LMC significantly outperforms state-of-the-art subgraph-wise sampling methods in terms of efficiency. | We propose a novel and efficient subgraph-wise sampling method with a convergence guarantee by Local Message Compensation (LMC). |
We present Branch-Train-Merge (BTM), a communication-efficient algorithm for embarrassingly parallel training of large language models (LLMs). We show it is possible to independently train subparts of a new class of LLMs on different subsets of the data, eliminating the massive multi-node synchronization currently required to train LLMs. BTM learns a set of independent Expert LMs (ELMs), each specialized to a different textual domain, such as scientific or legal text. These ELMs can be added and removed to update data coverage, ensembled to generalize to new domains, or averaged to collapse back to a single LM for efficient inference. New ELMs are learned by branching from (mixtures of) ELMs in the current set, further training on new domains, and then merging the resulting models back into the set for future use. Experiments show that BTM improves in- and out-of-domain perplexities as compared to GPT-style Transformer LMs, when controlling for training cost. Through extensive analysis, we show that these results are robust to different ELM initialization schemes, but require expert domain specialization; ensembles with random data splits do not perform well. Our results suggest that aggressive parallelism could be used to efficiently scale larger LMs in future work. | We develop a new class of large language models that is embarrassingly parallel: different parts of the model are independently trained on different subsets of the data, with no need for multi-node training or inference. |
Many important tasks involve some notion of long-term progress in multiple phases: e.g. to clean a shelf it must be cleared of items, cleaning products applied, and then the items placed back on the shelf. In this work, we explore the use of expert demonstrations in long-horizon tasks to learn a monotonically increasing function that summarizes progress. This function can then be used to aid agent exploration in environments with sparse rewards. As a case study we consider the NetHack environment, which requires long-term progress at a variety of scales and is far from being solved by existing approaches. In this environment, we demonstrate that by learning a model of long-term progress from expert data containing only observations, we can achieve efficient exploration in challenging sparse tasks, well beyond what is possible with current state-of-the-art approaches. We have made the curated gameplay dataset used in this work available at https://github.com/deepmind/nao_top10. | We learn a model of long-term progress using expert demonstrations, and show that it can be used to form an exploration reward that allows reinforcement learning agents to solve very challenging sparse tasks in NetHack. |
Ultra-High-Definition (UHD) photo has gradually become the standard configuration in advanced imaging devices. The new standard unveils many issues in existing approaches for low-light image enhancement (LLIE), especially in dealing with the intricate issue of joint luminance enhancement and noise removal while remaining efficient. Unlike existing methods that address the problem in the spatial domain, we propose a new solution, UHDFour, that embeds Fourier transform into a cascaded network. Our approach is motivated by a few unique characteristics in the Fourier domain: 1) most luminance information concentrates on amplitudes while noise is closely related to phases, and 2) a high-resolution image and its low-resolution version share similar amplitude patterns. Through embedding Fourier into our network, the amplitude and phase of a low-light image are separately processed to avoid amplifying noise when enhancing luminance. Besides, UHDFour is scalable to UHD images by implementing amplitude and phase enhancement under the low-resolution regime and then adjusting the high-resolution scale with few computations. We also contribute the first real UHD LLIE dataset, UHD-LL, that contains 2,150 low-noise/normal-clear 4K image pairs with diverse darkness and noise levels captured in different scenarios. With this dataset, we systematically analyze the performance of existing LLIE methods for processing UHD images and demonstrate the advantage of our solution. We believe our new framework, coupled with the dataset, would push the frontier of LLIE towards UHD. The code and dataset are available at https://li-chongyi.github.io/UHDFour/. | In this paper, we propose a new solution for UHD LLIE based on the characteristics of the Fourier domain. We also propose the first real UHD LLIE dataset with diverse data. |
This work introduces the Eigen Memory Tree (EMT), a novel online memory model for sequential learning scenarios. EMTs store data at the leaves of a binary tree, and route new samples through the structure using the principal components of previous experiences, facilitating efficient (logarithmic) access to relevant memories. We demonstrate that EMT outperforms existing online memory approaches, and provide a hybridized EMT-parametric algorithm that enjoys drastically improved performance over purely parametric methods with nearly no downsides. Our findings are validated using 206 datasets from OpenML repository in both bounded and infinite memory budget situations. | We create an episodic memory model for online learning and evaluate it for solving contextual bandit problems. |
Federated Learning (FL) is a machine learning paradigm where many clients collaboratively learn a shared global model with decentralized training data. Personalization on FL models additionally adapts the global model to different clients, achieving promising results on consistent local training & test distributions. However, for real-world personalized FL applications, it is crucial to go one step further: robustifying FL models under the evolving local test set during deployment, where various types of distribution shifts can arise. In this work, we identify the pitfalls of existing works under test-time distribution shifts and propose Federated Test-time Head Ensemble plus tuning (FedTHE+), which personalizes FL models with robustness to various test-time distribution shifts. We illustrate the advancement of FedTHE+ (and its degraded computationally efficient variant FedTHE) over strong competitors, for training various neural architectures (CNN, ResNet, and Transformer) on CIFAR10 and ImageNet and evaluating on diverse test distributions. Along with this, we build a benchmark for assessing the performance and robustness of personalized FL methods during deployment. Code: \url{https://github.com/LINs-lab/FedTHE}.
| We identify the pitfalls of existing personalized federated learning methods during deployment and propose a novel test-time personalization solution. |
The transfer learning paradigm of model pre-training and subsequent fine-tuning produces high accuracy models. However, a question remains: what data and method should be used for pre-training? We study the effect of the pre-training distribution on transfer learning in the context of image classification. Through controlled experiments, we find that the pre-training dataset is initially important for low-shot transfer. However, the differences between distributions is diminished as more data is made available for fine-tuning. Still, fine-tuning outperforms training from scratch. We also investigate dataset size and observe that larger pre-training datasets lead to better accuracy, however, the absolute accuracy difference is largest in the few-shot regime. Beyond data, we study the effect of the pre-training method, language-image contrastive vs. image-image contrastive, finding that the latter usually leads to better transfer accuracy | We investigate the role of pretraining distribution, data curation, size, and loss and downstream transfer learning |
Variational Autoencoder (VAE)-based generative models offer flexible representation learning by incorporating meta-priors, general premises considered beneficial for downstream tasks. However, the incorporated meta-priors often involve ad-hoc model deviations from the original likelihood architecture, causing undesirable changes in their training. In this paper, we propose a novel representation learning method, Gromov-Wasserstein Autoencoders (GWAE), which directly matches the latent and data distributions using the variational autoencoding scheme. Instead of likelihood-based objectives, GWAE models minimize the Gromov-Wasserstein (GW) metric between the trainable prior and given data distributions. The GW metric measures the distance structure-oriented discrepancy between distributions even with different dimensionalities, which provides a direct measure between the latent and data spaces. By restricting the prior family, we can introduce meta-priors into the latent space without changing their objective. The empirical comparisons with VAE-based models show that GWAE models work in two prominent meta-priors, disentanglement and clustering, with their GW objective unchanged. | GWAEs, our novel generative models, learn representations based on meta-priors by directly fitting their latent space into the data space. |
As the deployment of automated face recognition (FR) systems proliferates, bias in these systems is not just an academic question, but a matter of public concern. Media portrayals often center imbalance as the main source of bias, i.e., that FR models perform worse on images of non-white people or women because these demographic groups are underrepresented in training data. Recent academic research paints a more nuanced picture of this relationship. However, previous studies of data imbalance in FR have focused exclusively on the face verification setting, while the face identification setting has been largely ignored, despite being deployed in sensitive applications such as law enforcement. This is an unfortunate omission, as 'imbalance' is a more complex matter in identification; imbalance may arise in not only the training data, but also the testing data, and furthermore may affect the proportion of identities belonging to each demographic group or the number of images belonging to each identity. In this work, we address this gap in the research by thoroughly exploring the effects of each kind of imbalance possible in face identification, and discuss other factors which may impact bias in this setting. | In this work, we explore the effects of different kinds of data imbalance on bias in face identification problem. |
Deep Neural Networks are vulnerable to adversarial attacks, which makes adversarial attacks serve as a method to evaluate the robustness of DNNs. However, adversarial attacks have high white-box attack success rates but poor transferability, making black-box attacks impracticable in the real world. Momentum-based attacks were proposed to accelerate optimization to improve transferability. Nevertheless, conventional momentum-based attacks accelerate optimization inefficiently during early iterations since the initial value of momentum is zero, which leads to unsatisfactory transferability. Therefore, we propose Experienced Momentum (EM), which is the pre-trained momentum. Initializing the momentum to EM can help accelerate optimization during the early iterations. Moreover, the pre-update of conventional Nesterov momentum based attacks is rough, prompting us to propose Precise Nesterov momentum (PN). PN refines the pre-update by considering the gradient of the current data point. Finally, we integrate EM with PN as Experienced Precise Nesterov momentum (EPN) to further improve transferability. Extensive experiments against normally trained and defense models demonstrate that our EPN is more effective than conventional momentum in the improvement of transferability. Specifically, the attack success rates of our EPN-based attacks are $\sim$11.9% and $\sim$13.1% higher than conventional momentum-based attacks on average against normally trained and defense models, respectively. | Our proposed EPN is more effective than traditional momentum in improving transferability, and extensive experiments show that EPN-based attacks are more transferable than SOTA. |
The success of neural networks over the past decade has established them as effective models for many relevant data generating processes. Statistical theory on neural networks indicates graceful scaling of sample complexity. For example, Joen & Van Roy (An information-theoretic framework for supervised learning, 2022) demonstrate that, when data is generated by a ReLU teacher network with $W$ parameters, an optimal learner needs only $\tilde{O}(W/\epsilon)$ samples to attain expected error $\epsilon$. However, existing computational theory suggests that, even for single-hidden-layer teacher networks, to attain small error for all such teacher networks, the computation required to achieve this sample complexity is intractable. In this work, we fit single-hidden-layer neural networks to data generated by single-hidden-layer ReLU teacher networks with parameters drawn from a natural distribution. We demonstrate that stochastic gradient descent (SGD) with automated width selection attains small expected error with a number of samples and total number of queries both nearly linear in the input dimension and width. This suggests that SGD nearly achieves the information-theoretic sample complexity bounds of Joen & Van Roy (2022) in a computationally efficient manner. An important difference between our positive empirical results and the negative theoretical results is that the latter address worst-case error of deterministic algorithms, while our analysis centers on expected error of a stochastic algorithm. | We empirically demonstrate that SGD computationally efficiently achieves theoretical sample complexity bounds on single-hidden-layer teacher networks. |
Out-of-distribution (OOD) detection is an important task to ensure the reliability and safety of deep learning and the discriminator models outperform others for now. However, the feature extraction of such models must compress the data and lose certain information, leaving room for bad cases and malicious attacks. However, despite effectively fitting the data distribution and producing high-quality samples, generative models lack suitable indicator scores to match with discriminator models in the OOD detection tasks. In this paper, we find that these two kinds of models can be combined to solve each other's problems. We introduce diffusion models (DMs), a kind of powerful generative model, into OOD detection and find that the denoising process of DMs also functions as a novel form of asymmetric interpolation. This property establishes a diffusion-based neighborhood for each input data. Then, we perform discriminator-based OOD detection based on the diffusion-based neighborhood instead of isolated data. In this combination, the discriminator models provide detection metrics for generation models and the diffusion-based neighborhood reduces the information loss of feature extraction. According to our experiments on CIFAR10 and CIFAR100, our new methods successfully outperform state-of-the-art methods. Our implementation is put in the supplementary materials. | We design a general strategy to combine a diffusion model and a Resnet to do OOD detection. |
The offline reinforcement learning (RL) problem is often motivated by the need to learn data-driven decision policies in financial, legal and healthcare applications. However, the learned policy could retain sensitive information of individuals in the training data (e.g., treatment and outcome of patients), thus susceptible to various privacy risks. We design offline RL algorithms with differential privacy guarantees which provably prevent such risks. These algorithms also enjoy strong instance-dependent learning bounds under both tabular and linear Markov Decision Process (MDP) settings. Our theory and simulation suggest that the privacy guarantee comes at (almost) no drop in utility comparing to the non-private counterpart for a medium-size dataset. | We present the first provably efficient offline RL algorithms with differential privacy guarantees. |
The Neural Radiance Fields (NeRF) have been recently applied to reconstruct building-scale and even city-scale scenes. To model a large-scale scene efficiently, a dominant strategy is to employ a divide-and-conquer paradigm via performing scene decomposition, which decomposes a complex scene into parts that are further processed by different sub-networks. Existing large-scale NeRFs mainly use heuristic hand-crafted scene decomposition, with regular 3D-distance-based or physical-street-block-based schemes. Although achieving promising results, the hand-crafted schemes limit the capabilities of NeRF in large-scale scene modeling in several aspects. Manually designing a universal scene decomposition rule for different complex scenes is challenging, leading to adaptation issues for different scenarios. The decomposition procedure is not learnable, hindering the network from jointly optimizing the scene decomposition and the radiance fields in an end-to-end manner. The different sub-networks are typically optimized independently, and thus hand-crafted rules are required to composite them to achieve a better consistency. To tackle these issues, we propose Switch-NeRF, a novel end-to-end large-scale NeRF with learning-based scene decomposition. We design a gating network to dispatch 3D points to different NeRF sub-networks. The gating network can be optimized together with the NeRF sub-networks for different scene partitions, by a design with the Sparsely Gated Mixture of Experts (MoE). The outputs from different sub-networks can also be fused in a learnable way in the unified framework to effectively guarantee the consistency of the whole scene. Furthermore, the proposed MoE-based Switch-NeRF model is carefully implemented and optimized to achieve both high-fidelity scene reconstruction and efficient computation. Our method establishes clear state-of-the-art performances on several large-scale datasets. To the best of our knowledge, we are the first to propose an applicable end-to-end sparse NeRF network with learning-based decomposition for large-scale scenes. Codes are released at https://github.com/MiZhenxing/Switch-NeRF. | We propose an applicable end-to-end sparse NeRF network with learning-based decomposition for large-scale scenes. |
Graph neural networks have recently shown outstanding results in diverse types of tasks in machine learning, providing interdisciplinary state-of-the-art performance on structured data. However, they have been proved to be vulnerable to imperceptible adversarial attacks and shown to be unfit for out-of-distribution generalisation.
Here, we address this problem by introducing a novel message-passing scheme based on the theory of predictive coding, an energy-based alternative to back-propagation that has its roots in neuroscience.
As both graph convolution and predictive coding can be seen as low-pass filtering mechanisms, we postulate that predictive coding adds a second efficient filter to the messaging passing process which enhances the robustness of the learned representation. Through an extensive set of experiments, we show that the proposed model attains comparable performance to its graph convolution network counterpart, delivering strictly better performance on inductive tasks. Most importantly, we show that the energy minimization enhances the robustness of the produced presentation and can be leveraged to further calibrate our models and provide representations that are more robust against advanced graph adversarial attacks.
| For the first time, we use predictive coding in deep geometric learning and demonstrate that we can enhance the robustness of learning representation through energy minimization. |
In representation learning, a common approach is to seek representations which disentangle the underlying factors of variation. Eastwood & Williams (2018) proposed three metrics for quantifying the quality of such disentangled representations: disentanglement (D), completeness (C) and informativeness (I). In this work, we first connect this DCI framework to two common notions of linear and nonlinear identifiability, thereby establishing a formal link between disentanglement and the closely-related field of independent component analysis. We then propose an extended DCI-ES framework with two new measures of representation quality—explicitness (E) and size (S)—and point out how D and C can be computed for black-box predictors. Our main idea is that the functional capacity required to use a representation is an important but thus-far neglected aspect of representation quality, which we quantify using explicitness or ease-of-use (E). We illustrate the relevance of our extensions on the MPI3D and Cars3D datasets. | We extend the DCI framework for evaluating disentangled representations and connect it to identifiability. |
Scalability remains a challenge in multi-agent reinforcement learning and is currently under active research. However, existing works lack the ability to identify the essential interaction under the non-stationary environment. We propose causal mean field Q-learning (CMFQ) to address this problem. It has the advantage of MFQ, which can compress the space size dramatically. Besides, it is ever more robust toward the non-stationary caused by increasing agents. We enable agents to identify which ally or opponent is more crucial by asking "what if" with the help of the structural causal model (SCM), then pay more attention to more crucial ones. We test CMFQ in mixed cooperative-competitive and cooperative games, which verify our method's scalability performance. | This paper aims at the scalability problem in large-scale multi-agent system. We use causal inference to imporve the robustness of mean field Q-learning. Experiments verify that our method achieve superior scalability performance. |
Electronic health records (EHRs) typically contain a wide range of time series data that is characterized by high sparsity and irregular observations. Self-supervised Transformer architectures have shown outstanding performance in a variety of structured tasks in natural language processing and computer vision. However, their use in modelling sparse irregular time series with tabular data has not been widely explored. One of the major challenges is the quadratic scaling of self-attention layers that can significantly limit the input sequence length. In this work, we introduce TESS, Transformers for EHR data with Self Supervised learning, a self-supervised Transformer-based architecture designed to extract robust representations from EHR data. We propose an input binning scheme that aggregates the time series inputs and sparsity information into a regular sequence with fixed length, enabling the training of larger and deeper Transformers. We demonstrate that significant compression of EHR input data is possible without sacrificing useful information, likely due to the highly correlated nature of observations in small time bins. We then introduce self-supervised prediction tasks that provide rich and informative signals for model pre-training. TESS outperforms state-of-the-art deep learning models on multiple downstream tasks from the MIMIC-IV and PhysioNet-2012 EHR datasets. | We propose a Transformer based model for sparse time series that utilizes an input binning scheme to aggregate the time series inputs. |
Probabilistic forecasting of multivariate time series is significant to several research domains where multiple futures exist for a single observed sequence. Identifying the observations on which a well-performing model bases its forecasts can enable domain experts to form data-driven hypotheses about the causal relationships between features. Consequently, we begin by revisiting the question: what constitutes a causal explanation? One hurdle in the landscape of explainable artificial intelligence is that what constitutes an explanation is not well-grounded. We build upon Miller's framework of explanations derived from research in multiple social science disciplines, and establish a conceptual link between counterfactual reasoning and saliency-based explanation techniques. However, the complication is a lack of a consistent and principled notion of saliency. Also, commonly derived saliency maps may be inconsistent with the data generation process and the underlying model. We therefore leverage a unifying definition of information-theoretic saliency grounded in preattentive human visual cognition and extend it to forecasting settings. In contrast to existing methods that require either explicit training of the saliency mechanism or access to the internal parameters of the underlying model, we obtain a closed-form solution for the resulting saliency map for commonly used density functions in probabilistic forecasting. To empirically evaluate our explainability framework in a principled manner, we construct a synthetic dataset of conversation dynamics and demonstrate that our method recovers the true salient timesteps for a forecast given a well-performing underlying model. | We propose an information-theoretic saliency-based framework for counterfactual reasoning in probabilistic forecasting. For common distributions, we obtain a closed-form expression for the saliency of observed timesteps towards a model's forecasts. |
Practical results have shown that deep learning optimizers using small constant learning rates, hyperparameters close to one, and large batch sizes can find the model parameters of deep neural networks that minimize the loss functions. We first show theoretical evidence that the momentum method (Momentum) and adaptive moment estimation (Adam) perform well in the sense that the upper bound of the theoretical performance measure is small with a small constant learning rate, hyperparameters close to one, and a large batch size. Next, we show that there exists a batch size called the critical batch size minimizing the stochastic first-order oracle (SFO) complexity, which is the stochastic gradient computation cost, and that SFO complexity increases once the batch size exceeds the critical batch size. Finally, we provide numerical results that support our theoretical results. That is, the numerical results indicate that Adam using a small constant learning rate, hyperparameters close to one, and the critical batch size minimizing SFO complexity has faster convergence than Momentum and stochastic gradient descent (SGD). | Critical batch size minimizes stochastic first-order oracle complexity of deep learning optimizer using hyperparameters close to one. |
Unsupervised denoising is a crucial challenge in real-world imaging applications. Unsupervised deep-learning methods have demonstrated impressive performance on benchmarks based on synthetic noise. However, no metrics are available to evaluate these methods in an unsupervised fashion. This is highly problematic for the many practical applications where ground-truth clean images are not available. In this work, we propose two novel metrics: the unsupervised mean squared error (MSE) and the unsupervised peak signal-to-noise ratio (PSNR), which are computed using only noisy data. We provide a theoretical analysis of these metrics, showing that they are asymptotically consistent estimators of the supervised MSE and PSNR. Controlled numerical experiments with synthetic noise confirm that they provide accurate approximations in practice. We validate our approach on real-world data from two imaging modalities: videos in raw format and transmission electron microscopy. Our results demonstrate that the proposed metrics enable unsupervised evaluation of denoising methods based exclusively on noisy data. | We introduce two novel unsupervised metrics, uMSE and uPSNR, computed exclusively from noisy data, which are asymptotically consistent estimators of the corresponding supervised metrics, MSE and PSNR, and yield accurate approximations in practice |
Neural Algorithmic Reasoning is an emerging area of machine learning which seeks to infuse algorithmic computation in neural networks, typically by training neural models to approximate steps of classical algorithms. In this context, much of the current work has focused on learning reachability and shortest path graph algorithms, showing that joint learning on similar algorithms is beneficial for generalisation. However, when targeting more complex problems, such "similar" algorithms become more difficult to find. Here, we propose to learn algorithms by exploiting duality of the underlying algorithmic problem. Many algorithms solve optimisation problems. We demonstrate that simultaneously learning the dual definition of these optimisation problems in algorithmic learning allows for better learning and qualitatively better solutions. Specifically, we exploit the max-flow min-cut theorem to simultaneously learn these two algorithms over synthetically generated graphs, demonstrating the effectiveness of the proposed approach. We then validate the real-world utility of our dual algorithmic reasoner by deploying it on a challenging brain vessel classification task, which likely depends on the vessels’ flow properties. We demonstrate a clear performance gain when using our model within such a context, and empirically show that learning the max-flow and min-cut algorithms together is critical for achieving such a result. | A neural algorithmic reasoning approach exploiting the duality principle |
We propose global context vision transformer (GC ViT), a novel architecture that enhances parameter and compute utilization for computer vision tasks. The core of the novel model are global context self-attention modules, joint with standard local self-attention, to effectively yet efficiently model both long and short-range spatial interactions, as an alternative to complex operations such as an attention masks or local windows shifting. While the local self-attention modules are responsible for modeling short-range information, the global query tokens are shared across all global self-attention modules to interact with local key and values. In addition, we address the lack of inductive bias in ViTs and improve the modeling of inter-channel dependencies by proposing a novel downsampler which leverages a parameter-efficient fused inverted residual block. The proposed GC ViT achieves new state-of-the-art performance across image classification, object detection and semantic segmentation tasks. On ImageNet-1K dataset for classification, the tiny, small and base variants of GC ViT with 28M, 51M and 90M parameters achieve 83.4%, 83.9% and 84.4% Top-1 accuracy, respectively, surpassing comparably-sized prior art such as CNN-based ConvNeXt and ViT-based Swin Transformer. Pre-trained GC ViT backbones in downstream tasks of object detection, instance segmentation, and semantic segmentation on MS COCO and ADE20K datasets outperform prior work consistently, sometimes by large margins. | We introduce general computer vision backbone to effectively learn both short and long-range spatial information. |
The use of pretrained deep neural networks represents an attractive way to achieve strong results with few data available. When specialized in dense problems such as object detection, learning local rather than global information in images has proven to be more efficient. However, for unsupervised pretraining, the popular contrastive learning requires a large batch size and, therefore, a lot of resources. To address this problem, we are interested in transformer-based object detectors that have recently gained traction in the community with good performance and with the particularity of generating many diverse object proposals.
In this work, we present Proposal Selection Contrast (ProSeCo), a novel unsupervised overall pretraining approach that leverages this property. ProSeCo uses the large number of object proposals generated by the detector for contrastive learning, which allows the use of a smaller batch size, combined with object-level features to learn local information in the images. To improve the effectiveness of the contrastive loss, we introduce the object location information in the selection of positive examples to take into account multiple overlapping object proposals. When reusing pretrained backbone, we advocate for consistency in learning local information between the backbone and the detection head.
We show that our method outperforms state of the art in unsupervised pretraining for object detection on standard and novel benchmarks in learning with fewer data. | We present Proposal Selection Contrast (ProSeCo), a novel unsupervised overall pretraining approach for Object Detection that leverages the large number of object proposals generated by transformer-based detectors using an improved contrastive loss. |
At times hate speech detection classifiers miss the context of a sentence and flag a sarcastic tweet incorrectly. To tackle this problem by emphasising on the context of a tweet we propose SarNet. SarNet is a two-fold deep learning based model which follows a quasi-ternary labelling strategy and contextually classifies a tweet as hate, sarcastic or neither. The first module of SarNet is an ANN-BiLSTM based Pyramid Network used to calculate the hate and sarcastic probabilities of a sentence. The second module of the SarNet is the Nash Equalizer which stems from the concept of game theory and prisoner’s dilemma. It treats hate and sarcasm as two prisoners. A payoff matrix is constructed to calculate the true hate of the tweet. True hate considers the hate part of a tweet excluding the sarcastic part of the tweet. Thus, this gives a true estimate of the hate content in a tweet thereby decreasing the number of sarcastic tweets being falsely flagged as hate. Our proposed model is trained on state-of-the-art hate speech and sarcasm datasets in the English language. The precision, recall and F1 score of our proposed model is 0.93, 0.84 and 0.88 respectively. Comparison with state-of-the-art architectures demonstrated better performance of SarNet by a significant margin. | This research paper focuses on quasi-ternary classification of hate and sarcasm in a tweet using game theory, Nash Equilibrium and deep learning. |
Applications of deep neural networks are booming in more and more fields but lack transparency due to their black-box nature. Explainable Artificial Intelligence (XAI) is therefore of paramount importance, where strategies are proposed to understand how these black-box models function. The research so far mainly focuses on producing, for example, class-wise saliency maps, highlighting parts of a given image that affect the prediction the most. However, this way does not fully represent the way humans explain their reasoning and, awkwardly, validating these maps is quite complex and generally requires subjective interpretation. In this article, we conduct XAI differently by proposing a new XAI methodology in a multilevel (i.e., visual and linguistic) manner. By leveraging the interplay between the learned representations, i.e., image features and linguistic attributes, the proposed approach can provide salient attributes and attribute-wise saliency maps, which are far more intuitive than the class-wise maps, without requiring per-image ground-truth human explanations. It introduces self-interpretable attributes to overcome the current limitations in XAI and bring the XAI towards human-like level. The proposed architecture is simple in use and can reach surprisingly good performance in both prediction and explainability for deep neural networks thanks to the low-cost per-class attributes. | We propose a novel XAI methodology to explain DNNs predictions in a multilevel manner (i.e., visual and linguistic) without requiring per-image annotations. |
Simulating rigid collisions among arbitrary shapes is notoriously difficult due to complex geometry and the strong non-linearity of the interactions. While graph neural network (GNN)-based models are effective at learning to simulate complex physical dynamics, such as fluids, cloth and articulated bodies, they have been less effective and efficient on rigid-body physics, except with very simple shapes. Existing methods that model collisions through the meshes' nodes are often inaccurate because they struggle when collisions occur on faces far from nodes. Alternative approaches that represent the geometry densely with many particles are prohibitively expensive for complex shapes. Here we introduce the ``Face Interaction Graph Network'' (FIGNet) which extends beyond GNN-based methods, and computes interactions between mesh faces, rather than nodes. Compared to learned node- and particle-based methods, FIGNet is around 4x more accurate in simulating complex shape interactions, while also 8x more computationally efficient on sparse, rigid meshes. Moreover, FIGNet can learn frictional dynamics directly from real-world data, and can be more accurate than analytical solvers given modest amounts of training data. FIGNet represents a key step forward in one of the few remaining physical domains which have seen little competition from learned simulators, and offers allied fields such as robotics, graphics and mechanical design a new tool for simulation and model-based planning. | Face to face, multi-index collisions improve accuracy and efficiency of graph network models for rigid body dynamics |
The demand for large-scale computational resources for Neural Architecture Search (NAS) has been lessened by tabular benchmarks for NAS. Evaluating NAS strategies is now possible on extensive search spaces and at a moderate computational cost. But so far, NAS has mainly focused on maximising performance on some hold-out validation/test set. However, energy consumption is a partially conflicting objective that should not be neglected. We hypothesise that constraining NAS to include the energy consumption of training the models could reveal a sub-space of undiscovered architectures that are more computationally efficient with a smaller carbon footprint. To support the hypothesis, an existing tabular benchmark for NAS is augmented with the energy consumption of each architecture. We then perform multi-objective optimisation that includes energy consumption as an additional objective. We demonstrate the usefulness of multi-objective NAS for uncovering the trade-off between performance and energy consumption as well as for finding more energy-efficient architectures. The updated tabular benchmark is open-sourced to encourage the further exploration of energy consumption-aware NAS. | Energy consumption-aware tabular benchmarks for NAS can be used to access sub-space of architectures that are inherently efficient. |
Asymmetric image retrieval aims to deploy compatible models on platforms of different resources to achieve a balance between computational efficiency and retrieval accuracy. The most critical issue is how to align the output features of different models. Despite the great progress, existing approaches apply strong constraints so that features or neighbor structures are strictly aligned across different models. However, such a one-to-one constraint is too strict to be well preserved for the query models with low capacity. Considering that the primary concern of the users is the rank of the returned images, we propose a generic rank preserving framework, which achieves feature compatibility and the order consistency between query and gallery models simultaneously. Specifically, we propose two alternatives to instantiate the framework. One realizes straightforward rank order preservation by directly preserving the consistency of the sorting results. To make sorting process differentiable, the Heaviside step function in sorting is approximated by the sigmoid function. The other aims to preserve a learnable monotonic mapping relationship between the returned similarity scores of query and gallery models. The mapped similarity scores of gallery model are considered as pseudo-supervision to guide the query model training. Extensive experiments on various large-scale datasets demonstrate the superiority of our two proposed methods. | We propose a rank preserving framework to achieve the consistency of the ranking lists returned by asymmetric and symmetric retrieval. |
The current approach to ML model design is either to choose a flexible Blackbox model and explain it post hoc or to start with an interpretable model. Blackbox models are flexible but difficult to explain, whereas interpretable models are designed to be explainable. However, developing interpretable models necessitates extensive ML knowledge, and the resulting models tend to be less flexible, offering potentially subpar performance compared to their Blackbox equivalents. This paper aims to blur the distinction between a post hoc explanation of a BlackBox and constructing interpretable models. We propose beginning with a flexible BlackBox model and gradually carving out a mixture of interpretable models and a residual network. Our design identifies a subset of samples and routes them through the interpretable models. The remaining samples are routed through a flexible residual network. We adopt First Order Logic (FOL) as the interpretable model's backbone, which provides basic reasoning on the concept retrieved from the BlackBox model. On the residual network, we repeat the method until the proportion of data explained by the residual network falls below a desired threshold.
Our approach offers several advantages. First, the mixture of interpretable and flexible residual networks results in almost no compromise in performance. Second, the rout, interpret, and repeat approach yields a highly flexible interpretable model. Our extensive experiment demonstrates the performance of the model on various datasets. We show that by editing the FOL model, we can fix the shortcut learned by the original BlackBox model. Finally, our method provides a framework for a hybrid symbolic-connectionist network that is simple to train and adaptable to many applications. | We seek to carve out the interpretable model from trained blackbox iteratively and explain the prediction in terms of human interpretable concepts using First order logic (FOL)) |
The ability to interpret machine learning models is critical for high-stakes applications. Due to its desirable theoretical properties, path integration is a widely used scheme for feature attribution to interpret model predictions. However, the methods implementing this scheme currently rely on absolute attribution scores to eventually provide sensible interpretations. This not only contradicts the premise that the features with larger attribution scores are more relevant to the model prediction, but also conflicts with the theoretical settings for which the desirable properties of the attributions are proven. We address this by devising a method to first compute an appropriate reference for the path integration scheme. This reference further helps in identifying valid interpolation points on a desired integration path. The reference is computed in a gradient ascending direction on the model's loss surface, while the interpolations are performed by analyzing the model gradients and variations between the reference and the input. The eventual integration is effectively performed along a non-linear path. Our scheme can be incorporated into the existing integral-based attribution methods. We also devise an effective sampling and integration procedure that enables employing our scheme with multi-reference path integration efficiently. We achieve a marked performance boost for a range of integral-based attribution methods on both local and global evaluation metrics by enhancing them with our scheme. Our extensive results also show improved sensitivity, sanity preservation and model robustness with the proposed re-calibration of the attribution techniques with our method. | We propose a re-calibration technique to calibrate existing integral-based attribution methods with valid references for a consistent explanation. |
In recent years, a proliferation of methods were developed for cooperative multi-agent reinforcement learning (c-MARL). However, the robustness of c-MARL agents against adversarial attacks has been rarely explored. In this paper, we propose to evaluate the robustness of c-MARL agents via a model-based approach, named c-MBA. Our proposed formulation can craft much stronger adversarial state perturbations of c-MARL agents to lower total team rewards than existing model-free approaches. In addition, we propose the first victim-agent selection strategy and the first data-driven approach to define targeted failure states where each of them allows us to develop even stronger adversarial attack without the expert knowledge to the underlying environment. Our numerical experiments on two representative MARL benchmarks illustrate the advantage of our approach over other baselines: our model-based attack consistently outperforms other baselines in all tested environments. | A novel model-based adversarial attack framework for cooperative multi-agent reinforcement learning with novel victim agent selection strategy. |
Large-scale Pre-Trained Language Models (PTLMs) capture knowledge from massive human-written data which contains latent societal biases and toxic contents. In this paper, we leverage the primary task of PTLMs, i.e. language modeling, and propose a new metric to quantify manifested implicit representational harms in PTLMs towards 13 marginalized demographics. Using this metric, we conducted an empirical analysis of 24 widely used PTLMs. Our analysis provides insights into the correlation between the proposed metric in this work and other related fairness metrics. We observe that our metric correlates with the majority of gender-specific fairness metrics in the literature. Through extensive experiments, we explore the connections between PTLMs architectures and representational harms across two dimensions: depth and width of the networks. We found that prioritizing depth over width, mitigates representational harms in some PTLMs. | Measuring implicit hate in pretrained language models |
Deep neural networks are notorious for defying theoretical treatment. However, when the number of parameters in each layer tends to infinity the network function is a Gaussian process (GP) and quantitatively predictive description is possible. Gaussian approximation allows to formulate criteria for selecting hyperparameters, such as variances of weights and biases, as well as the learning rate. These criteria rely on the notion of criticality defined for deep neural networks. In this work we describe a new practical way to diagnose criticality. We introduce *partial Jacobians* of a network, defined as derivatives of preactivations in layer $l$ with respect to preactivations in layer $l_0\leq l$. We derive recurrence relations for the norms of partial Jacobians and utilize these relations to analyze criticality of deep fully connected neural networks with LayerNorm and/or residual connections. We derive and implement a simple and cheap numerical test that allows one to select optimal initialization for a broad class of deep neural networks. Using these tools we show quantitatively that proper stacking of the LayerNorm (applied to preactivations) and residual connections leads to an architecture that is critical for any initialization. Finally, we apply our methods to analyze the MLP-Mixer architecture and show that it is everywhere critical. | We introduce a new diagnostic for critical initialization in deep neural networks; and show that a combination of LayerNorm and residual connections leads to everywhere critical architectures. |
We propose $\textit{Masked Encoding for Tabular Data (MET)}$ for learning self-supervised representations from $\textit{tabular data}$. Tabular self-supervised learning (tabular-SSL) -- unlike structured domains like images, audio, text -- is more challenging since each tabular dataset can have a completely different structure among its features (or coordinates), which is hard to identify a priori. $\textit{MET}$ attempts to circumvent this problem by assuming the following hypothesis: the observed tabular data features come from a latent graphical model and the downstream tasks are significantly easier to solve in the latent space. Based on this hypothesis, $\textit{MET}$ uses random masking based encoders to learn a positional embedding for each coordinate, which would in turn capture the latent structure between coordinates. Through experiments on a toy dataset from a linear graphical model, we show that $\textit{MET}$ is indeed able to capture the latent graphical model. Practically, through extensive experiments on multiple benchmarks for tabular data, we demonstrate that $\textit{MET}$ significantly outperforms all the baselines. For example, on Criteo -- a large-scale click prediction dataset -- $\textit{MET}$ achieves as much as $5\%$ improvement over the current state-of-the-art (SOTA) while purely supervised learning based approaches have been able to advance SOTA by at most $2\%$ in the last six years. Furthermore, averaged over $\textit{nine}$ datasets, $\textit{MET}$ is around $3.9\%$ more accurate than the next best method of Gradient-boosted decision trees -- considered as SOTA for the tabular setting. | Masking based algorithm for SSL on tabular datasets. Key idea: there exists a latent graphical model that captures relations between different coordinates and classification in latent space is easy. Masking based SSL learns this latent structure. |
In off-policy reinforcement learning, an agent collects transition data (a.k.a. experience tuples) from the environment and stores them in a replay buffer for the incoming parameter updates. Storing those tuples consumes a large amount of memory when the environment observations are given as images. Large memory consumption is especially problematic when reinforcement learning methods are applied in scenarios where the computational resources are limited. In this paper, we introduce a method to prune relatively unimportant experience tuples by a simple metric that estimates the importance of experiences and saves the overall memory consumption by the buffer. To measure the importance of experiences, we use $\textit{surprise}$ and $\textit{on-policyness}$. Surprise is quantified by the information gain the model can obtain from the experiences and on-policyness ensures that they are relevant to the current policy. In our experiments, we empirically show that our method can significantly reduce the memory consumption by the replay buffer without decreasing the performance in vision-based environments. | We propose a method to prune experiences in the replay buffer using a metric based on surprise and on-policyness of the experience and use it to save memory consumption in off-policy reinforcement learning. |
The goal of multi-objective reinforcement learning (MORL) is to learn policies that simultaneously optimize multiple competing objectives. In practice, an agent's preferences over the objectives may not be known apriori, and hence, we require policies that can generalize to arbitrary preferences at test time. In this work, we propose a new data-driven setup for offline MORL, where we wish to learn a preference-agnostic policy agent using only a finite dataset of offline demonstrations of other agents and their preferences. The key contributions of this work are two-fold. First, we introduce D4MORL, (D)atasets for MORL that are specifically designed for offline settings. It contains 1.8 million annotated demonstrations obtained by rolling out reference policies that optimize for randomly sampled preferences on 6 MuJoCo environments with 2-3 objectives each. Second, we propose Pareto-Efficient Decision Agents (PEDA), a family of offline MORL algorithms that builds and extends Decision Transformers via a novel preference-and-return-conditioned policy. Empirically, we show that PEDA closely approximates the behavioral policy on the D4MORL benchmark and provides an excellent approximation of the Pareto-front with appropriate conditioning, as measured by the hypervolume and sparsity metrics. | We introduce new dataset & benchmarks and propose new algorithms for offline Multi-Objective Reinforcement Learning (MORL) |