abstract
stringlengths 383
2.62k
| TLDR
stringlengths 4
250
|
|---|---|
Transfer learning is known to perform efficiently in many applications empirically, yet limited literature reports the mechanism behind the scene. This study establishes both formal derivations and heuristic analysis to formulate the theory of transfer learning in deep learning. Our framework utilizing layer variational analysis proves that the success of transfer learning can be guaranteed with corresponding data conditions. Moreover, our theoretical calculation yields intuitive interpretations towards the knowledge transfer process. Subsequently, an alternative method for network-based transfer learning is derived. The method shows an increase in efficiency and accuracy for domain adaptation. It is particularly advantageous when new domain data is sufficiently sparse during adaptation. Numerical experiments over diverse tasks validated our theory and verified that our analytic expression achieved better performance in domain adaptation than the gradient descent method. | Interpretations of Domain Adaptations via Layer Variational Analysis |
For many interdisciplinary fields,
ML interpretations need to be consistent with \emph{what-if} scenarios related to the current case, i.e., if one factor changes, how does the model react?
Although the attribution methods are supported by the elegant axiomatic systems, they mainly focus on individual inputs,
and are generally inconsistent.
To support what-if scenarios, we introduce a new objective of consistency based on a notion called truthful interpretation. Towards this objective,
we apply Fourier analysis of Boolean functions to get
consistency guarantees.
Experimental results show that
for neighborhoods with various radii,
our method achieves $2$x - $50$x lower inconsistency compared with the other methods. | We find that the previous attribution methods are not consistent with neighborhood predictions, and introduce a new framework with an efficient algorithm to support consistency. |
Semi-supervised learning and weakly supervised learning are important paradigms that aim to reduce the growing demand for labeled data in current machine learning applications. In this paper, we introduce a novel analysis of the classical label propagation algorithm (LPA) (Zhu & Ghahramani, 2002) that moreover takes advantage of useful prior information, specifically probabilistic hypothesized labels on the unlabeled data. We provide an error bound that exploits both the local geometric properties of the underlying graph and the quality of the prior information. We also propose a framework to incorporate multiple sources of noisy information. In particular, we consider the setting of weak supervision, where our sources of information are weak labelers. We demonstrate the ability of our approach on multiple benchmark weakly supervised classification tasks, showing improvements upon existing semi-supervised and weakly supervised methods. | Theoretical analysis of label propagation with prior information and connection to weak supervision |
A major challenge in reinforcement learning is specifying tasks in a manner that is both interpretable and verifiable. One common approach is to specify tasks through reward machines---finite state machines that encode the task to be solved. We introduce skill machines, a representation that can be learned directly from these reward machines that encode the solution to such tasks. We propose a framework where an agent first learns a set of base skills in a reward-free setting, and then combines these skills with the learned skill machine to produce composite behaviours specified by any regular language, such as linear temporal logics. This provides the agent with the ability to map from complex logical task specifications to near-optimal behaviours zero-shot. We demonstrate our approach in both a tabular and high-dimensional video game environment, where an agent is faced with several of these complex, long-horizon tasks. Our results indicate that the agent is capable of satisfying extremely complex task specifications, producing near optimal performance with no further learning. Finally, we demonstrate that the performance of skill machines can be improved with regular off-policy reinforcement learning algorithms when optimal behaviours are desired. | A framework where an agent first learns a set of base skills in a reward-free setting, and then combines them with the learned skill machine to produce composite behaviours specified by any regular language, such as linear temporal logics. |
Survival analysis is the problem of estimating probability distributions for future events, which can be seen as a problem in uncertainty quantification. Although there are fundamental theories on strictly proper scoring rules for uncertainty quantification, little is known about those for survival analysis. In this paper, we investigate extensions of four major strictly proper scoring rules for survival analysis. Through the extensions, we discuss and clarify the assumptions arising from the discretization of the estimation of probability distributions. We also discuss the relationship between the existing algorithms and extended scoring rules, and we propose new algorithms based on our extensions of the scoring rules for survival analysis.
| Theoretical analysis of scoring rules for survival analysis. |
Artificial neural networks (ANNs) exhibit a narrow scope of expertise on stationary independent data. However, data in the real world is continuous and dynamic, and ANNs must adapt to novel scenarios while also retaining the learned knowledge to become lifelong learners. The ability of humans to excel at these tasks can be attributed to multiple factors ranging from cognitive computational structures, cognitive biases, and the multi-memory systems in the brain. We incorporate key concepts from each of these to design a cognitive-inspired continual learning method. Cognitive Continual Learner (CCL) includes multiple modules, implicit and explicit knowledge representation dichotomy, inductive bias, and a multi-memory system. CCL shows improvement across all continual learning settings and also exhibits reduced task recency bias. To test versatility of continual learning methods on a challenging distribution shift, we introduce a novel domain-incremental dataset Domain${^2}$IL. In addition to improved performance on existing benchmarks, CCL also demonstrates superior performance on this dataset. | Cognitive Continual Learner (CCL), a novel cognitive-inspired method that employs multiple modules, implicit and explicit knowledge representation dichotomy, inductive bias, and a multi-memory system. |
Systems consisting of interacting agents are prevalent in the world, ranging from dynamical systems in physics to complex biological networks. To build systems which can interact robustly in the real world, it is thus important to be able to infer the precise interactions governing such systems. Existing approaches typically discover such interactions by explicitly modeling the feedforward dynamics of the trajectories. In this work, we propose Neural Constraint Inference (NCI) model as an alternative approach to discover such interactions: it discovers a set of relational constraints, represented as energy functions, which when optimized reconstruct the original trajectory. We illustrate how NCI can faithfully predict future trajectory dynamics, achieving more consistent long-rollouts than existing approaches. We show that the constraints discovered by NCI are disentangled and may be intermixed with constraints from other trajectories. Finally, we illustrate how those constraints enable the incorporation of external test-time constraints. | We propose an approach that discovers a set of relational constraints, represented as energy functions, which when optimized reconstruct a given original trajectory. |
By enabling agents to communicate, recent cooperative multi-agent reinforcement learning (MARL) methods have demonstrated better task performance and more coordinated behavior. Most existing approaches facilitate inter-agent communication by allowing agents to send messages to each other through free communication channels, i.e., \emph{cheap talk channels}. Current methods require these channels to be constantly accessible and known to the agents a priori. In this work, we lift these requirements such that the agents must discover the cheap talk channels and learn how to use them. Hence, the problem has two main parts: \emph{cheap talk discovery} (CTD) and \emph{cheap talk utilization} (CTU). We introduce a novel conceptual framework for both parts and develop a new algorithm based on mutual information maximization that outperforms existing algorithms in CTD/CTU settings. We also release a novel benchmark suite to stimulate future research in CTD/CTU. | A novel problem formulation and methodology in MARL on learning where to communicate and where best to communicate. |
In real-world scenarios, subgraphs of a larger global graph may be distributed across multiple devices or institutions, and only locally accessible due to privacy restrictions, although there may be links between them. Recently proposed subgraph Federated Learning (FL) methods deal with those missing links across private local subgraphs while distributively training Graph Neural Networks (GNNs) on them. However, they have overlooked the inevitable heterogeneity among subgraphs, caused by subgraphs comprising different communities of a global graph, therefore, consequently collapsing the incompatible knowledge from local GNN models trained on heterogeneous graph distributions. To overcome such a limitation, we introduce a new subgraph FL problem, personalized subgraph FL, which focuses on the joint improvement of the interrelated local GNN models rather than learning a single global GNN model, and propose a novel framework, FEDerated Personalized sUBgraph learning (FED-PUB), to tackle it. A crucial challenge in personalized subgraph FL is that the server does not know which subgraph each client has. FED-PUB thus utilizes functional embeddings of the local GNNs using random graphs as inputs to compute similarities between them, and use them to perform weighted averaging for server-side aggregation. Further, it learns a personalized sparse mask at each client to select and update only the subgraph-relevant subset of the aggregated parameters. We validate FED-PUB for its subgraph FL performance on six datasets, considering both non-overlapping and overlapping subgraphs, on which ours largely outperforms relevant baselines. | A novel personalized subgraph federated learning framework aiming at the joint improvement of interrelated local models trained on interconnected local subgraphs, for instance, subgraphs belonging to the same community. |
We prove that image classifiers are fundamentally sensitive to small perturbations in their inputs. Specifically, we show that given some image space of $n$-by-$n$ images, all but a tiny fraction of images in any image class induced over that space can be moved outside that class by adding some perturbation whose $p$-norm is $O(n^{1/\max{(p,1)}})$, as long as that image class takes up at most half of the image space. We then show that $O(n^{1/\max{(p,1)}})$ is asymptotically optimal. Finally, we show that an increase in the bit depth of the image space leads to a loss in robustness. We supplement our results with a discussion of their implications for vision systems. | Image classifiers are fundamentally sensitive to small perturbations in their inputs. |
Federated learning (FL) has recently gained considerable attention due to its ability to learn on decentralised data while preserving client privacy. However, it also poses additional challenges related to the heterogeneity of the participating devices, both in terms of their computational capabilities and contributed data. Meanwhile, Neural Architecture Search (NAS) has been successfully used with centralised datasets, producing state-of-the-art results in constrained or unconstrained settings. However, such centralised datasets may not be always available for training. Most recent work at the intersection of NAS and FL attempts to alleviate this issue in a cross-silo federated setting, which assumes homogeneous compute environments with datacenter-grade hardware.
In this paper we explore the question of whether we can design architectures of different footprints in a cross-device federated setting, where the device landscape, availability and scale are very different. To this end, we design our system, FedorAS, to discover and train promising architectures in a resource-aware manner when dealing with devices of varying capabilities holding non-IID distributed data. We present empirical evidence of its effectiveness across different settings, spanning across three different modalities (vision, speech, text), and showcase its better performance compared to state-of-the-art federated solutions, while maintaining resource efficiency. | FedorAS is a system that performs cross-device Federated Neural Architecture Search under heterogeneous system and data distributions. |
Convolutional models have been widely used in multiple domains. However, most existing models only use local convolution, making the model unable to handle long-range dependencies efficiently. Attention overcomes this problem by aggregating global information based on the pair-wise attention score but also makes the computational complexity quadratic to the sequence length. Recently, Gu et al. proposed a model called S4 inspired by the state space model. S4 can be efficiently implemented as a global convolutional model whose kernel size equals the input sequence length. With Fast Fourier Transform, S4 can model much longer sequences than Transformers and achieve significant gains over SoTA on several long-range tasks. Despite its empirical success, S4 is involved. It requires sophisticated parameterization and initialization schemes that combine the wisdom from several prior works. As a result, S4 is less intuitive and hard to use for researchers with limited prior knowledge. Here we aim to demystify S4 and extract basic principles that contribute to the success of S4 as a global convolutional model. We focus on the structure of the convolution kernel and identify two critical but intuitive principles enjoyed by S4 that are sufficient to make up an effective global convolutional model: 1) The parameterization of the convolutional kernel needs to be efficient in the sense that the number of parameters should scale sub-linearly with sequence length. 2) The kernel needs to satisfy a decaying structure that the weights for convolving with closer neighbors are larger than the more distant ones. Based on the two principles, we propose a simple yet effective convolutional model called Structured Global Convolution (SGConv). SGConv exhibits strong empirical performance over several tasks: 1) With faster speed, SGConv surpasses the previous SoTA on Long Range Arena and Speech Command datasets. 2) When plugging SGConv into standard language and vision models, it shows the potential to improve both efficiency and performance. | We proposed a simple Strucured Global Convolution Kernel for long-range dependencies. |
This paper revisits datasets and evaluation criteria for Symbolic Regression, a task of expressing given data using mathematical equations, specifically focused on its potential for scientific discovery. Focused on a set of formulas used in the existing datasets based on Feynman Lectures on Physics, we recreate 120 datasets to discuss the performance of symbolic regression for scientific discovery (SRSD). For each of the 120 SRSD datasets, we carefully review the properties of the formula and its variables to design reasonably realistic sampling range of values so that our new SRSD datasets can be used for evaluating the potential of SRSD such as whether or not an SR method can (re)discover physical laws from such datasets. As an evaluation metric, we also propose to use normalized edit distances between a predicted equation and the ground-truth equation trees. While existing metrics are either binary or errors between the target values and an SR model's predicted values for a given input, normalized edit distances evaluate a sort of similarity between the ground-truth and predicted equation trees. We have conducted experiments on our new SRSD datasets using five state-of-the-art SR methods in SRBench and a simple baseline based on a recent Transformer architecture. The results show that we provide a more realistic performance evaluation and open up a new machine learning-based approach for scientific discovery. We provide our datasets and code as part of the supplementary material. | We propose new datasets and evaluation metric to discuss the performance of symbolic regression for scientific discovery (SRSD). |
Distributed computing has been a promising solution in machine learning to accelerate the training procedure on large-scale dataset by utilizing multiple workers in parallel. However, there remain two major issues that still need to be addressed: i) adversarial attacks from malicious workers, and ii) the effect of slow workers known as stragglers. In this paper, we tackle both problems simultaneously by proposing Group-wise Verifiable Coded Computing (GVCC), which leverages coding techniques and group-wise verification to provide robustness to adversarial attacks and resiliency to straggler effects in distributed computing. The key idea of GVCC is to verify a group of computation results from workers at a time, while providing resilience to stragglers through encoding tasks assigned to workers with Group-wise Verifiable Codes. Experimental results show that GVCC outperforms the existing methods in terms of overall processing time and verification time for executing matrix multiplication, which is a key computational component in machine learning and deep learning. | This paper tackles adversarial attack and straggler effect in distributed computing by proposing Group-wise Verifiable Coded Computing. |
When deploying Reinforcement Learning (RL) agents into a physical system, we must ensure that these agents are well aware of the underlying constraints. In many real-world problems, however, the constraints are often hard to specify mathematically and unknown to the RL agents. To tackle these issues, Inverse Constrained Reinforcement Learning (ICRL) empirically estimates constraints from expert demonstrations. As an emerging research topic, ICRL does not have common benchmarks, and previous works tested algorithms under hand-crafted environments with manually-generated expert demonstrations. In this paper, we construct an ICRL benchmark in the context of RL application domains, including robot control, and autonomous driving. For each environment, we design relevant constraints and train expert agents to generate demonstration data. Besides, unlike existing baselines that learn a deterministic constraint, we propose a variational ICRL method to model a posterior distribution of candidate constraints. We conduct extensive experiments on these algorithms under our benchmark and show how they can facilitate studying important research challenges for ICRL. The benchmark, including the instructions for reproducing ICRL algorithms, is available at https://github.com/Guiliang/ICRL-benchmarks-public. | We design a benchmark with important applications for Inverse Constrained Reinforcement Learning and propose a variational Bayesian approach for modeling the distribution of constraints. |
Offline reinforcement learning (RL) has attracted a great deal of attention recently as an approach to utilizing past experience to learn a policy. Recent studies have reported the challenges of offline RL, such as estimating the values of actions that are out of the data distribution. To mitigate the issues of offline RL, we propose an algorithm that leverages a mixture of deterministic policies. With our framework, the state-action space is divided by learning discrete latent variables, and sub-policies corresponding to each region are trained. The proposed algorithm, which we call Value-Weighted Variational Auto-Encoder (V2AE), is derived by considering the variational lower bound of the offline RL objective function. The aim of this work is to shed lights on the importance on the policy structure in offline RL. We show empirically that the use of the proposed mixture policy can reduce the accumulation of the approximation error in offline RL, which was reported in previous studies. Experimental results also indicate that introducing the policy structure improves the performance on tasks with D4RL benchmarking datasets. | We introduce a structure in a policy representation in offline reinforcement learning, which reduces the critic loss during the training and improves the resulting policy performance. |
A variety of pruning methods have been introduced for over-parameterized recurrent neural networks to improve efficiency in terms of power and storage. With the advance in pruning methods and their variety, a new problem of ‘hyperpruning’ is becoming apparent: finding a suitable pruning method with optimal hyperparameter configuration for a particular task and network. Such search is different from the standard hyperparameter search, where the accuracy of the optimal configuration is unknown. In the context of network pruning, the accuracy of the non-pruned (dense) model sets the target for the accuracy of the pruned model. Thereby, the goal of hyperpruning is to reach or even surpass this target. It is critical to develop efficient strategies for hyperpruning since direct search through pruned variants would require time-consuming training without guarantees for improved performance. To address this problem, we introduce a novel distance based on Lyapunov Spectrum (LS) which provides means to compare pruned variants with the dense model and early in training to estimate the accuracy that pruned variants will achieve after extensive training. The ability to predict performance allows us to incorporate the LS-based distance with Bayesian hyperparameter optimization methods and to propose an efficient and first-of-its-kind hyperpruning approach called LS-based Hyperpruning (LSH) which can optimize the search time by an order of magnitude compared to standard full training search with the loss (or perplexity) being the accuracy metric. Our experiments on stacked LSTM and RHN language models trained with the Penn Treebank dataset show that with a given budget of training epochs and desired pruning ratio, LSH obtains more optimal variants than standard loss-based hyperparameter optimization methods. Furthermore, as a result of the search, LSH identifies pruned variants that outperform state-of-the-art pruning methods and surpass the accuracy of the dense model. | We proposed a novel method to search over pruning method and hyperparameters based on Lyapunov Spectrum. |
We address the problem of group fairness in classification, where the objective is to learn models that do not unjustly discriminate against subgroups of the population. Most existing approaches are limited to simple binary tasks or involve difficult to implement training mechanisms. This reduces their practical applicability. In this paper, we propose FairGrad, a method to enforce fairness based on a reweighting scheme that iteratively learns group specific weights based on whether they are advantaged or not. FairGrad is easy to implement and can accommodate various standard fairness definitions. Furthermore, we show that it is competitive with standard baselines over various datasets including ones used in natural language processing and computer vision. | A method to enforce fairness based on a reweighting scheme that iteratively learns group specific weights based on whether they are advantaged or not. |
In this work, we tackle the problem of Open-Ended Learning by a method that simultaneously evolves agents and increasingly challenging environments. Unlike previous open-ended approaches that optimize agents using a fixed neural network topology, we hypothesize that generalization can be improved by allowing agents' controllers to become more complex as they encounter more difficult environments. Our method, Augmentative Topology EPOET (ATEP), extends the Enhanced Paired Open-Ended Trailblazer (EPOET) algorithm by allowing agents to evolve their own neural network structures over time, adding complexity and capacity as necessary. Empirical results demonstrate that ATEP results in general agents capable of solving more environments than a fixed-topology baseline. We also investigate mechanisms for transferring agents between environments and find that a species-based approach further improves the performance and generalization of agents. | This work brings generalization capabilities and ability to solve complex environments to Open Ended Learning framework by adding agents that augment their topologies over time. |
We present a general framework for evolutionary learning to emergent unbiased state representation without any supervision. Evolutionary frameworks such as self-play converge to bad local optima in case of multi-agent reinforcement learning in non-cooperative partially observable environments with communication due to information asymmetry. Our proposed framework is a simple modification of self-play inspired by mechanism design, also known as {\em reverse game theory}, to elicit truthful signals and make the agents cooperative. The key idea is to add imaginary rewards using the peer prediction method, i.e., a mechanism for evaluating the validity of information exchanged between agents in a decentralized environment. Numerical experiments with predator prey, traffic junction and StarCraft tasks demonstrate that the state-of-the-art performance of our framework. | TSP is a general framework for evolutionary learning to emergent unbiased state representation without any supervision. |
Deep neural networks are powerful tools for representation learning, but can easily overfit to noisy labels which are prevalent in many real-world scenarios. Generally, noisy supervision could stem from variation among labelers, label corruption by adversaries, etc. To combat such label noises, one popular line of approach is to apply customized weights to the training instances, so that the corrupted examples contribute less to the model learning. However, such learning mechanisms potentially erase important information about the data distribution and therefore yield suboptimal results. To leverage useful information from the corrupted instances, an alternative is the bootstrapping loss, which reconstructs new training targets on-the-fly by reweighting the real labels and the network's own predictions (i.e., pseudo labels).
In this paper, we propose a more generic learnable loss objective which enables a joint reweighting of instances and labels at once. Specifically, our method dynamically adjusts the $\textit{per-sample importance weight}$ between the real observed labels and pseudo-labels, where the weights are efficiently determined in a meta process. Compared to the previous instance reweighting methods, our approach concurrently conducts implicit relabeling, and thereby yields substantial improvements with almost no extra cost. Extensive experimental results demonstrated the strengths of our approach over existing methods on multiple natural and medical image benchmark datasets, including CIFAR-10, CIFAR-100, ISIC2019 and Clothing 1M. Code will be made publicly available. | A simple and effective method for combating the label noise via joint instance and label reweighting |
In this work, we explore the maximum-margin bias of quasi-homogeneous neural networks trained with gradient flow on an exponential loss and past a point of separability. We introduce the class of quasi-homogeneous models, which is expressive enough to describe nearly all neural networks with homogeneous activations, even those with biases, residual connections, and normalization layers, while structured enough to enable geometric analysis of its gradient dynamics. Using this analysis, we generalize the existing results of maximum-margin bias for homogeneous networks to this richer class of models. We find that gradient flow implicitly favors a subset of the parameters, unlike in the case of a homogeneous model where all parameters are treated equally. We demonstrate through simple examples how this strong favoritism toward minimizing an asymmetric norm can degrade the robustness of quasi-homogeneous models. On the other hand, we conjecture that this norm-minimization discards, when possible, unnecessary higher-order parameters, reducing the model to a sparser parameterization. Lastly, by applying our theorem to sufficiently expressive neural networks with normalization layers, we reveal a universal mechanism behind the empirical phenomenon of Neural Collapse. | We generalize implicit max-margin bias to a class of models which describes nearly all networks, identifying a competition between maximizing margin and minimizing an asymmetric parameter norm, which can degrade robustness and explain Neural Collapse |
In this paper, we focus on a typical two-phase phenomenon in the learning of multi-layer perceptrons (MLPs). We discover and explain the reason for the feature collapse phenomenon in the first phase, i.e., the diversity of features over different samples keeps decreasing in the first phase, until samples of different categories share almost the same feature, which hurts the optimization of MLPs. We explain such a phenomenon in terms of the learning dynamics of MLPs. Furthermore, we theoretically analyze the reason why four typical operations can alleviate the feature collapse. The code has been attached with the submission. | In this paper, we focus on a typical two-phase phenomenon in the learning of multi-layer perceptrons (MLPs), and we discover and explain the reason for the feature collapse in the first phase. |
Denoising generative models (DGMs) have recently attracted increasing attention by showing impressive synthesis quality. DGMs are built on a diffusion process that pushes data to the noise distribution and the models learn to denoise. In this paper, we establish the interpretation of DGMs in terms of image restoration (IR). Integrating IR literature allows us to use an alternative objective and diverse forward processes, not confining to the diffusion process. By imposing prior knowledge on the loss function grounded on MAP estimation, we eliminate the need for the expensive sampling of DGMs. Also, we propose a multi-scale training, which alleviates the latent inefficiency of DGMs, by taking advantage of the flexibility of the forward process. Our model improves the quality and efficiency of both training and inference, achieving state-of-the-art performance when the number of forward steps is limited. Furthermore, we show the applicability of our model to inverse problems. We believe that our framework paves the way for designing a new type of flexible general generative model. | A new framework on generative modeling in the perspective of restoration. |
We propose DoE2Vec, a variational autoencoder (VAE)-based methodology to learn optimization landscape characteristics for downstream meta-learning tasks, e.g., automated selection of optimization algorithms. Principally, using large training data sets generated with a random function generator, DoE2Vec self-learns an informative latent representation for any design of experiments (DoE).
Unlike the classical exploratory landscape analysis (ELA) method, our approach does not require any feature engineering and is easily applicable for high dimensional search spaces. For validation, we inspect the quality of latent reconstructions and analyze the latent representations using different experiments.
The latent representations not only show promising potentials in identifying similar (cheap-to-evaluate) surrogate functions, but also can boost performances when being used complementary to the ELA features in classification tasks. | We propose DoE2Vec, a variational autoencoder (VAE)-based methodology to learn optimization landscape characteristics for downstream meta-learning tasks. |
We address the problem of learning geometric representations from observations perceived by an agent operating within an environment and interacting with an external object. To this end, we propose a representation learning framework that extracts the state of both the agent and the object from unstructured observations of arbitrary nature (e.g., images). Supervision comes from the performed actions alone, while the dynamics of the object is assumed to be unknown. We provide a theoretical foundation and formally prove that an ideal learner is guaranteed to infer an isometric representation, disentangling the agent from the object. Finally, we investigate empirically our framework on a variety of scenarios. Results show that our model reliably infers the correct representation and outperforms vision-based approaches such as a state-of-the-art keypoint extractor.
| We propose a representation learning framework that extracts from observations the geometric state of both an agent and an object the agent interacts with. |
In the age of Bigdata, Federated Learning (FL) provides machine learning (ML) practitioners with an indispensable tool for solving large-scale learning problems. FL is a distributed optimization paradigm where multiple nodes each having access to a local dataset collaborate (with or without a server) to solve a joint problem. Federated Averaging (FedAvg) although the algorithm of choice for many FL applications is not very well understood especially in the interpolation regime, a common phenomenon observed in modern overparameterized neural networks. In this work, we address this challenge and perform a thorough theoretical performance analysis of FedAvg in the interpolation regime for training of overparameterized neural networks. Specifically, we analyze the performance of FedAvg in two settings: (i) {\em[Server]}: When the network has access to a server that coordinates the information sharing among nodes, and (ii) {\em[Decentralized]:} The serverless setting, where the local nodes communicate over an undirected graph. We consider a class of non-convex functions satisfying the Polyak-Lojasiewicz (PL) condition, a condition that is satisfied by overparameterized neural networks. For the first time, we establish that FedAvg under both {\em Server} and {\em Decentralized} settings achieve linear convergence rates of $\mathcal{O}(T^{3/2} \log (1/{\epsilon} ) )$ and $\mathcal{O}({T^2} \log ({1}/{\epsilon}))$, respectively, where $\epsilon$ is the desired solution accuracy, and $T$ is the number of local updates at each node. In contrast to the standard FedAvg analysis, our work does not require bounded heterogeneity, variance, and gradient assumptions. Instead, we show that sample-wise (and local) smoothness of the local loss functions suffice to capture the effect of heterogeneity in FL training. We use a novel application of induction to prove the linear convergence in the {\em Decentralized} setting, which can be of independent interest. Finally, we conduct experiments on multiple real datasets to corroborate our theoretical findings. | Our work shows linear convergence for Federated Averaging algorithm in {\em Server} and {\em Decentralized} settings. |
Offline reinforcement learning (RL), which aims to learn an optimal policy using a previously collected static dataset, is an important paradigm of RL. Standard RL methods often perform poorly in this regime due to the function approximation errors on out-of-distribution actions. While a variety of regularization methods have been proposed to mitigate this issue, they are often constrained by policy classes with limited expressiveness that can lead to highly suboptimal solutions. In this paper, we propose representing the policy as a diffusion model, a recent class of highly-expressive deep generative models. We introduce Diffusion Q-learning (Diffusion-QL) that utilizes a conditional diffusion model to represent the policy. In our approach, we learn an action-value function and we add a term maximizing action-values into the training loss of the conditional diffusion model, which results in a loss that seeks optimal actions that are near the behavior policy. We show the expressiveness of the diffusion model-based policy, and the coupling of the behavior cloning and policy improvement under the diffusion model both contribute to the outstanding performance of Diffusion-QL. We illustrate the superiority of our method compared to prior works in a simple 2D bandit example with a multimodal behavior policy. We then show that our method can achieve state-of-the-art performance on the majority of the D4RL benchmark tasks. | Diffusion models serve as expressive policies to boost offline RL performance. |
Reliable application of machine learning-based decision systems in the wild is one of the major challenges currently investigated by the field. A large portion of established approaches aims to detect erroneous predictions by means of assigning confidence scores. This confidence may be obtained by either quantifying the model's predictive uncertainty, learning explicit scoring functions, or assessing whether the input is in line with the training distribution. Curiously, while these approaches all state to address the same eventual goal of detecting failures of a classifier upon real-world application, they currently constitute largely separated research fields with individual evaluation protocols, which either exclude a substantial part of relevant methods or ignore large parts of relevant failure sources. In this work, we systematically reveal current pitfalls caused by these inconsistencies and derive requirements for a holistic and realistic evaluation of failure detection. To demonstrate the relevance of this unified perspective, we present a large-scale empirical study for the first time enabling benchmarking confidence scoring functions w.r.t all relevant methods and failure sources. The revelation of a simple softmax response baseline as the overall best performing method underlines the drastic shortcomings of current evaluation in the plethora of publicized research on confidence scoring. Code and trained models are at https://github.com/https://github.com/IML-DKFZ/fd-shifts | We present a holistic perspective on the task of failure detection including a large-scale empirical study for the first time enabling benchmarking confidence scoring functions w.r.t all relevant methods and distribution shifts. |
Graph Neural Networks (GNNs) gained popularity to address the tasks over the graph-structured data that best represent many real-world systems. The privacy of the participants of these systems is at risk if the GNNs are not carefully designed. Existing works in privacy-preserving GNNs primarily ensure the privacy of features and labels of a node. In order to ensure complete privacy related to graph data, its structure also needs to be privatized. We provide a method SPGraph to privatize the graph structure by adding noise to the neighborhood data of the node. Our method addresses two challenges in introducing structural privacy in graphs. Applying randomization on the set of actual neighbors to introduce noise leads to a reduction in the degree of a node, which is undesirable. To overcome this first challenge, we introduce $\lambda$-selector that samples nodes to be added to the set of neighbors. The second challenge is to denoise the neighborhood so that the noise added in the neighborhood does not significantly impact the accuracy. In this view, we use $p$-hop neighborhood to compensate for the loss of actual neighbors in the randomization. We continue to use the node and label privacy as implemented in the previous methods for privacy in GNNs. We conduct extensive experiments over real-world datasets to show the impact of perturbation in the graph structure. | Make the structure of the graph private in addition to the privacy of node features and labels |
Probabilistic Graphical Models are generative models of complex systems. They rely on conditional independence assumptions between variables to learn sparse representations which can be visualized in a form of a graph. Such models are used for domain exploration and structure discovery in poorly understood domains. This work introduces a novel technique to perform sparse graph recovery by optimizing deep unrolled networks. Assuming that the input data $X\in\mathbb{R}^{M\times D}$ comes from an underlying multivariate Gaussian distribution, we apply a deep model on $X$ that outputs the precision matrix $\Theta$. Then, the partial correlation matrix \mathrm{P} is calculated which can also be interpreted as providing a list of conditional independence assertions holding in the input distribution. Our model, \texttt{uGLAD}, builds upon and extends the state-of-the-art model \texttt{GLAD} to the unsupervised setting. The key benefits of our model are (1) \texttt{uGLAD} automatically optimizes sparsity-related regularization parameters leading to better performance than existing algorithms. (2) We introduce multi-task learning based `consensus' strategy for robust handling of missing data in an unsupervised setting. We evaluate performance on synthetic Gaussian, non-Gaussian data generated from Gene Regulatory Networks, and present case studies in anaerobic digestion and infant mortality. | An unsupervised deep learning method based on unrolled algorithm technique to recover conditional independence graphs. |
We study the problem of compositional generalization of language-instructed agents in gSCAN. gSCAN is a popular benchmark which requires an agent to generalize to instructions containing novel combinations of words, which are not seen in the training data. We propose to improve the agent’s generalization capabilities with an architecture inspired by the Meta-Sequence-to-Sequence learning approach (Lake, 2019). The agent receives as a context a few examples of pairs of instructions and action trajectories in a given instance of the environment (a support set) and it is tasked to predict an action sequence for a query instruction for the same environment instance. The context is generated by an oracle and the instructions come from the same distribution as seen in the training data. In each training episode, we also shuffle the indices of the actions and the words of the instructions to make the agent figure out the relations between the actions and the words from the context. Our predictive model has the standard transformer architecture. We show that the proposed architecture can significantly improve the generalization capabilities of the agent on one of the most difficult gSCAN splits: the ``adverb-to-verb” split H. | We extend meta-seq2seq to grounded environments by generating environment relevant meta-training supports. |
By representing knowledge in a primary triple associated with additional attribute value qualifiers, hyper-relational knowledge graph (HKG) that generalizes triple based knowledge graph (KG) has been attracting research attention recently. Compared with KG, HKG is enriched with the semantic difference between the primary triple and additional qualifiers as well as the structural connection between entities in hyper-relational graph structure. However, to model HKG, existing studies mainly focus on either semantic information or structural information therein, fail to capture both simultaneously. To tackle this issue, in this paper, we propose an equivalent transformation for HKG modeling, referred to as TransEQ. Specifically, the equivalent transformation transforms a HKG to a KG, which considers both semantic and structural characteristics. Then a generalized encoder-decoder framework is developed to bridge the modeling research between KG and HKG. In the encoder part, KG-based graph neural networks are leveraged for structural modeling; while in the decoder part, various HKG-based scoring functions are exploited for semantic modeling. Especially, we design the sharing embedding mechanism in the encoder-decoder framework with semantic relatedness captured. We further theoretically prove that TransEQ preserves complete information in the equivalent transformation, and also achieves full expressivity. Finally, extensive experiments on three benchmarks demonstrate the superior performance of TransEQ in terms of both effectiveness and efficiency. On the largest benchmark WikiPeople, TransEQ significantly improves the state-of-the-art models by 15% on MRR. | We propose a simple yet effective transformation strategy for hyper-relational knowledge graph modeling with both semantic and structural information captured. |
Predictive performance of machine learning models trained with empirical risk minimization (ERM) can degrade considerably under distribution shifts. In particular, the presence of spurious correlations in training datasets leads ERM-trained models to display high loss when evaluated on minority groups not presenting such correlations in test sets. Extensive attempts have been made to develop methods improving worst-group robustness. However, they require group information for each training input or at least, a validation set with group labels to tune their hyperparameters, which may be expensive to get or unknown a priori. In this paper, we address the challenge of improving group robustness without group annotations during training. To this end, we propose to partition automatically the training dataset into groups based on Gram matrices of features extracted from an identification model and to apply robust optimization based on these pseudo-groups. In the realistic context where no group labels are available, our experiments show that our approach not only improves group robustness over ERM but also outperforms all recent baselines. | We improve group robustness without group annotations by introducing GramClust, a two-stage method which (1) partition a dataset into groups based on features Gram matrices and (2) apply a robust optimization based on these pseudo-groups. |
A large body of research in continual learning is devoted to overcoming the catastrophic forgetting of neural networks by designing new algorithms that are robust to the distribution shifts. However, the majority of these works are strictly focused on the "algorithmic" part of continual learning for a "fixed neural network architecture", and the implications of using different architectures are not clearly understood. The few existing continual learning methods that expand the model also assume a fixed architecture and develop algorithms that can efficiently use the model throughout the learning experience. In contrast, in this work, we build on existing works that study continual learning from a neural network's architecture perspective and provide new insights into how the architecture choice, for the same learning algorithm, can impact stability-plasticity trade-off resulting in markedly different continual learning performance. We empirically analyze the impact of various architectural components providing best practices and recommendations that can improve the continual learning performance irrespective of the learning algorithm. | The choice of architecture can significantly impact the continual learning performance. |
Compositional representations of the world are a promising step towards enabling high-level scene understanding and efficient transfer to downstream tasks. Learning such representations for complex scenes and tasks remains an open challenge. Towards this goal, we introduce Neural Radiance Field Codebooks (NRC), a scalable method for learning object-centric representations through novel view reconstruction. NRC learns to reconstruct scenes from novel views using a dictionary of object codes which are decoded through a volumetric renderer. This enables the discovery of reoccurring visual and geometric patterns across scenes which are transferable to downstream tasks. We show that NRC representations transfer well to object navigation in THOR, outperforming 2D and 3D representation learning methods by 3.1\% success rate. We demonstrate that our approach is able to perform unsupervised segmentation for more complex synthetic (THOR) and real scenes (NYU Depth) better than prior methods (.101 ARI). Finally, we show that NRC improves on the task of depth ordering by 5.5% accuracy in THOR. | Learning geometrically-aware, object-centric representations through elastic bottlenecks and a differentiable renderer for downstream tasks. |
Visual imitation learning enables reinforcement learning agents to learn to behave from expert visual demonstrations such as videos or image sequences, without explicit, well-defined rewards.
Previous reseaches either adopt supervised learning techniques or induce simple and coarse scalar rewards from pixels, neglecting the dense information contained in the image demonstrations.
In this work, we propose to measure the expertise of various local regions of image samples, or called patches, and recover multi-dimensional patch rewards accordingly.
Patch reward is a more precise rewarding characterization that serves as fine-grained expertise measurement and visual explainability tool.
Specifically, we present Adversarial Imitation Learning with Patch Rewards (PatchAIL), which employs a patch-based discriminator to measure the expertise of different local parts from given images and provide patch rewards.
The patch-based knowledge is also used to regularize the aggregated reward and stabilize the training.
We evaluate our method on the standard pixel-based benchmark DeepMind Control Suite.
The experiment results have demonstrated that PatchAIL outperforms baseline methods and provides valuable interpretations for visual demonstrations. | We leverage to learn patch reward and present PatchAIL, an intuitive and principled learning framework for efficient visual imitation learning. |
A natural goal in multi-agent learning is to learn \emph{rationalizable} behavior, where players learn to avoid any Iteratively Dominated Action (IDA). However, standard no-regret based equilibria-finding algorithms could take exponential samples to find such rationalizable strategies. In this paper, we first propose a simple yet sample-efficient algorithm for finding a rationalizable action profile in multi-player general-sum games under bandit feedback, which substantially improves over the results of Wu et al. We further develop algorithms with the first efficient guarantees for learning rationalizable Coarse Correlated Equilibria (CCE) and Correlated Equilibria (CE). Our algorithms incorporate several novel techniques to guarantee the elimination of IDA and no (swap-)regret simultaneously, including a correlated exploration scheme and adaptive learning rates, which may be of independent interest. We complement our results with a sample complexity lower bound showing the sharpness of our guarantees. | We develop provably efficient algorithms for finding approximate CE and CCE that are also rationalizable. |
Point clouds are an increasingly common spatial data modality, being produced by sensors used in robotics and self-driving cars, and as natural intermediate representations of objects in microscopy and other bioimaging domains (e.g., cell locations over time, or filaments, membranes, or organelle boundaries in cryo-electron micrographs or tomograms). However, semantic and instance segmentation of this data remains challenging due to the complex nature of objects in point clouds. Especially in bioimaging domains where objects are often large and can be intersecting or overlapping. Furthermore, methods for operating on point clouds should not be sensitive to the specific orientation or translation of the point cloud, which is often arbitrary. Here, we frame the point cloud instance segmentation problem as a graph learning problem in which we seek to learn a function that accepts the point cloud as input and outputs a probability distribution over neighbor graphs in which connected components of the graph correspond to individual object instances. We introduce the Dimensionless Instance Segmentation Transformer (DIST), a deep neural network for spatially invariant instance segmentation of point clouds to solve this point cloud-to-graph problem. DIST uses an SO(n) invariant transformer layer architecture to operate on point clouds of arbitrary dimension and outputs, for each pair of points, the probability that an edge exists between them in the instance graph. We then decode the most likely set of instances using a graph cut. We demonstrate the power of DIST for the segmentation of biomolecules in cryo-electron micrographs and tomograms, far surpassing existing methods for membrane and filament segmentation in empirical evaluation. DIST also applies to scene and object understanding, performing competitively on the ScanNetV2 3D instance segmentation challenge. We anticipate that DIST will underpin a new generation of methods for point cloud segmentation in bioimaging and that our general model and approach will provide useful insights for point cloud segmentation methods in other domains. | Novel method for learning point cloud graph representation for instance segmentation |
Many real-world problems like modelling environment dynamics, physical processes, time series etc., involve solving Partial Differential Equations (PDEs) parameterized by problem-specific conditions. Recently, a deep learning architecture called Fourier Neural Operator (FNO) proved to be capable of learning solutions of given PDE families, for any initial conditions as input. Given the advancements in quantum hardware and the recent results in quantum machine learning methods, we propose three quantum circuits, inspired by the FNO, to learn this functional mapping for PDEs. The proposed algorithms are distinguished based on the trade-off between depth and their similarity to the classical FNO. At their core, we make use of unary encoding paradigm and orthogonal quantum layers, and introduce a new quantum Fourier transform in the unary basis. With respect to the number of samples, our quantum algorithm is proven to be substantially faster than the classical counterpart. We benchmark our proposed algorithms on three PDE families, namely Burger’s equation, Darcy’s flow equation and the Navier-Stokes equation, and the results show that our quantum methods are comparable in performance to the classical FNO. We also show an analysis of the image classification tasks where our proposed algorithms are able to match the accuracy of the CNNs, thereby showing their applicability to other domains. | We provide three new quantum circuits to reproduce the Fourier Neural Operator, in order to learn PDEs solutions, and tested them on practical use cases. |
We present ReMasker, a novel method for imputing missing values in tabular data by extending the masked autoencoding framework. In contrast to prior work, ReMasker is both {\em simple} -- besides the missing values (i.e., naturally masked), we randomly ``re-mask'' another set of values, optimize the autoencoder by reconstructing this re-masked set, and apply the trained model to predict the missing values; and {\em effective} -- with extensive evaluation on benchmark datasets, we show that ReMasker consistently outperforms state-of-the-art methods in terms of both imputation fidelity and utility under various missingness settings, while its performance advantage often increases with the ratio of missing data. We further explore theoretical justification for its effectiveness, showing that ReMasker tends to learn missingness-invariant representations of tabular data. Our findings indicate that masked modeling represents a promising direction for further research on tabular data imputation. | We present ReMasker, an extremely simple yet effective method for imputing missing values in tabular data. |
Scientific Machine Learning (SciML) designs machine learning methods that predict physical systems governed by partial differential equations (PDE). These ML-based surrogate models substitute inefficient and often non-differentiable numerical simulation algorithms and find multiple applications such as weather forecasting, molecular dynamics, and medical applications.
While a number of ML-based methods for approximating the solutions of PDEs have been proposed in recent years, they typically do not consider the parameters of the PDEs, making it difficult for the ML surrogate models to generalize to PDE parameters not seen during training.
We propose a new channel-attention-based parameter embedding (CAPE) component for scientific machine learning models and a simple and effective curriculum learning strategy. The CAPE module can be combined with any kind of ML surrogate model, which can adapt to changing PDE parameters without harming the original model's ability to find approximate solutions to PDEs. The curriculum learning strategy provides a seamless transition between teacher-forcing and fully auto-regressive training.
We compare CAPE in conjunction with the curriculum learning strategy using a PDE benchmark and obtain consistent and significant improvements over the base models. The experiments also show several advantages of CAPE, such as its increased ability to generalize to unseen PDE parameters without substantially increasing inference time and parameter count.
An implementation of the method and experiments are available at \url{https://anonymous.4open.science/r/CAPE-ML4Sci-145B}. | a new parameter embedding module based on channel-attention for scientific machine learning |
We consider the problem of scheduling operations/nodes, the dependency among which is characterized by a Directed Acyclic Graph (DAG). Due to its NP-hard nature, heuristic algorithms were traditionally used to acquire reasonably good solutions, and more recent works have proposed Machine Learning (ML) heuristics that can generalize to unseen graphs and outperform the non-ML heuristics. However, it is computationally costly to generate solutions using existing ML schedulers since they adopt the episodic reinforcement learning framework that necessitates multi-round neural network processing. We propose a novel ML scheduler that uses a one-shot neural network encoder to sample node priorities which are converted by list scheduling to the final schedules. Since the one-shot encoder can efficiently sample the priorities in parallel, our algorithm runs significantly faster than existing ML baselines and has comparable run time with the fast traditional heuristics. We empirically show that our algorithm generates better schedules than both non-neural and neural baselines across various real-world and synthetic scheduling tasks. | We propose a novel ML scheduler that uses a one-shot neural network encoder to sample node priorities which are converted by list scheduling to the final schedules. |
A recourse action aims to explain a particular algorithmic decision by showing one specific way in which the instance could be modified to receive an alternate outcome. Existing recourse generation methods often assume that the machine learning model does not change over time. However, this assumption does not always hold in practice because of data distribution shifts, and in this case, the recourse action may become invalid. To redress this shortcoming, we propose the Distributionally Robust Recourse Action (DiRRAc) framework, which generates a recourse action that has high probability of being valid under a mixture of model shifts. We first formulate the robustified recourse setup as a min-max optimization problem, where the max problem is specified by Gelbrich distance over an ambiguity set around the distribution of model parameters. Then we suggest a projected gradient descent algorithm to find a robust recourse according to the min-max objective. We also show that our DiRRAc framework can be extended to hedge against the misspecification of the mixture weights. Numerical experiments with both synthetic and three real-world datasets demonstrate the benefits of our proposed framework over the state-of-the-art recourse methods, which generate robust recourses.
| Distributionally Robust Recourse Action framework generates a recourse action that has high probability of being valid under a mixture of model shifts. |
We study the task of prompting large-scale language models to perform multi-step reasoning. Existing work shows that when prompted with a chain of thoughts (CoT), sequences of short sentences describing intermediate reasoning steps towards a final answer, large language models can generate new reasoning chains and predict answers for new inputs. A central question is which reasoning examples make the most effective prompts. In this work, we propose complexity-based prompting, a simple and effective example selection scheme for multi-step reasoning. We show that prompts with higher reasoning complexity, i.e., chains with more reasoning steps, achieve substantially better performance on math word reasoning tasks over strong baselines. We further extend our complexity-based criteria from prompting (selecting inputs) to decoding (selecting outputs), where we sample multiple reasoning chains from the model, then choose the majority
of generated answers from complex reasoning chains (over simple chains). When used to prompt GPT-3, our approach substantially improves multi-step reasoning accuracy, with an 8.6% absolute improvement on GSM8K, and 6.4% on MathQA. Compared with existing example selection schemes like manual tuning or retrieval-based selection, selection based on reasoning complexity is intuitive, easy to implement, and annotation-efficient. Further results demonstrate the robustness of performance gains from complex prompts under format perturbation and distribution shift. | We show using prompts with more reasoning steps can improve language models multi-step reasoning ability |
In the era of deep learning, transferring information from a pretrained network to a downstream task by finetuning has many benefits. The choice of task head plays an important role in fine-tuning, as the pretrained and downstream tasks are usually different. Although there exist many different designs for finetuning, a full understanding of when and why these algorithms work has been elusive. We analyze how the choice of task head controls feature adaptation and hence influences the downstream performance. By decomposing the feature's learning dynamics, we find the key aspect is the training accuracy and loss at the beginning of finetuning, which determines the "energy" available for the feature's adaptation. We identify a significant trend in the effect of changes in this initial energy on the resulting features after finetuning. Specifically, as the energy increases, the Euclidean and cosine distances between the resulting and original features increase, while their dot product (and the resulting features’ norm) first increases and then decreases. Inspired by this, we give several practical principles that lead to better downstream performance. We analytically prove this trend in an overparamterized linear setting and verify its applicability to different experimental settings. | Features need mild adaptation during finetuning, so mildly update your task head and then finetune together. |
Retrosynthetic planning plays a critical role in drug discovery and organic chemistry. Starting from a target molecule as the root node, it aims to find a complete reaction tree subject to the constraint that all leaf nodes belong to a set of starting materials. The multi-step reactions are crucial because they determine the flow chart in the production of the Organic Chemical Industry. However, existing datasets lack curation of tree-structured multi-step reactions and fail to provide such reaction trees, limiting models' understanding of organic molecule transformations. In this work, we first develop a benchmark curated for the retrosynthetic planning task, which consists of 124,869 reaction trees retrieved from the public USPTO-full dataset. On top of that, we propose Metro: Memory-Enhanced Transformer for RetrOsynthetic planning. Specifically, the dependency among molecules in the reaction tree is captured as context information for multi-step retrosynthesis predictions through transformers with a memory module. Extensive experiments show that Metro dramatically outperforms existing single-step retrosynthesis models by at least 10.7% in top-1 accuracy. The experiments demonstrate the superiority of exploiting context information in the retrosynthetic planning task. Moreover, the proposed model can be directly used for synthetic accessibility analysis, as it is trained on reaction trees with the shortest depths. Our work is the first step towards a brand new formulation for retrosynthetic planning in the aspects of data construction, model design, and evaluation. | We use reaction database to search the retrosynthetic routes and introduce a memory network to learn the context information of the route. |
When training a variational autoencoder (VAE) on a given dataset, determining the optimal number of latent variables is mostly done by grid search: a costly process in terms of computational time and carbon footprint.
In this paper, we explore the intrinsic dimension estimation (IDE) of the data and latent representations learned by VAEs.
We show that the discrepancies between the IDE of the mean and sampled representations of a VAE after only a few steps of training reveal the presence of passive variables in the latent space, which, in well-behaved VAEs, indicates a superfluous number of dimensions.
Using this property, we propose FONDUE: an algorithm which quickly finds the number of latent dimensions after which the mean and sampled representations start to diverge (i.e., when passive variables are introduced), providing a principled method for selecting the number of latent dimensions for VAEs and autoencoders. | A principled method using intrinsic dimension estimation to find the optimal number of latent dimensions for variational autoencoders. |
Federated learning is a learning paradigm that allows the central server to learn from different data sources while keeping the data private at local. Without controlling and monitoring the local data collection process, it is highly likely that the locally available training labels are noisy, just as in a centralized data collection effort. Moreover, different clients may hold samples within different label spaces. The noisy label space is likely to be different from the unobservable clean label space, resulting in openset noisy labels. In this work, we study the challenge of federated learning from clients with openset noisy labels. We observe that many existing solutions, e.g., loss correction, in the noisy label literature cannot achieve their originally claimed effect in local training. A central contribution of this work is to propose an approach that communicates globally randomly selected ``contrastive labels" among clients to prevent local models from memorizing the openset noise patterns individually. Randomized label generations are applied during label sharing to facilitate access to the contrastive labels while ensuring differential privacy (DP). Both the DP guarantee and the effectiveness of our approach are theoretically guaranteed. Compared with several baseline methods, our solution shows its efficiency in several public benchmarks and real-world datasets under different noise ratios and noise models. | A framework for openset noisy label classification in federated learning |
Humans outperform object recognizers despite the fact that models perform well on current datasets. Numerous attempts have been made to create more challenging datasets by scaling them up from the web, exploring distribution shift, or adding controls for biases. The difficulty of each image in each dataset is not independently evaluated, nor is the concept of dataset difficulty as a whole well-posed. We develop a new dataset difficulty metric based on how long humans must view an image in order to classify a target object. Images whose objects can be recognized in 17ms are considered to be easier than those which require seconds of viewing time. Using 133,588 judgments on two major datasets, ImageNet and ObjectNet, we determine the distribution of image difficulties in those datasets, which we find varies wildly, but significantly undersamples hard images. Rather than hoping that distribution shift or other approaches will lead to hard datasets, we should measure the difficulty of datasets and seek to explicitly fill out the class of difficult examples. Analyzing model performance guided by image difficulty reveals that models tend to have lower performance and a larger generalization gap on harder images. Encouragingly for the biological validity of current architectures, much of the variance in human difficulty can be accounted for given an object recognizer by computing a combination of prediction depth, c-score, and adversarial robustness. We release a dataset of such judgments as a complementary metric to raw performance and a network’s ability to explain neural recordings. Such experiments with humans allow us to create a metric for progress in object recognition datasets, which we find are skewed toward easy examples, to test the biological validity of models in a novel way, and to develop tools for shaping datasets as they are being gathered to focus them on filling out the missing class of hard examples from today’s datasets. Dataset and analysis code can be found at https://github.com/image-flash/image-flash-2022. | We develop a new dataset difficulty metric based on how long humans must view an image in order to classify a target object, finding that the distribution of current datasets is skewed towards easy images. |
As data becomes increasingly vital, a company would be very cautious about releasing data, because the competitors could use it to train high-performance models, thereby posing a tremendous threat to the company's commercial competence. To prevent training good models on the data, we could add imperceptible perturbations to it. Since such perturbations aim at hurting the entire training process, they should reflect the vulnerability of DNN training, rather than that of a single model. Based on this new idea, we seek perturbed examples that are always unrecognized (never correctly classified) in training. In this paper, we uncover them by model checkpoints' gradients, forming the proposed self-ensemble protection (SEP), which is very effective because (1) learning on examples ignored during normal training tends to yield DNNs ignoring normal examples; (2) checkpoints' cross-model gradients are close to orthogonal, meaning that they are as diverse as DNNs with different architectures. That is, our amazing performance of ensemble only requires the computation of training one model. By extensive experiments with 9 baselines on 3 datasets and 5 architectures, SEP is verified to be a new state-of-the-art, e.g., our small $\ell_\infty=2/255$ perturbations reduce the accuracy of a CIFAR-10 ResNet18 from 94.56% to 14.68%, compared to 41.35% by the best-known method. Code is available at https://github.com/Sizhe-Chen/SEP. | We protect proprietary datasets by using intermediate checkpoints in a self-ensemble way, which more than halves the testing accuracy in unauthorized training compared to the best baselines. |
Contrastive learning has recently achieved remarkable success in many domains including graphs. However contrastive loss, especially for graphs, requires a large number of negative samples which is unscalable and computationally prohibitive with a quadratic time complexity. Sub-sampling is not optimal and incorrect negative sampling leads to sampling bias. In this work, we propose a meta-node based approximation technique that can (a) proxy all negative combinations (b) in quadratic cluster size time complexity, (c) at graph level, not node level, and (d) exploit graph sparsity. By replacing node-pairs with additive cluster-pairs, we compute the negatives in cluster-time at graph level. The resulting Proxy approximated meta-node Contrastive (PamC) loss, based on simple optimized GPU operations, captures the full set of negatives, yet is efficient with a linear time complexity. By avoiding sampling, we effectively eliminate sample bias. We meet the criterion for larger number of samples, thus achieving block-contrastiveness, which is proven to outperform pair-wise losses. We use learnt soft cluster assignments for the meta-node constriction, and avoid possible heterophily and noise added during edge creation. Theoretically, we show that real world graphs easily satisfy conditions necessary for our approximation. Empirically, we show promising accuracy gains over state-of-the-art graph clustering on 6 benchmarks. Importantly, we gain substantially in efficiency; up to 3x in training time, 1.8x in inference time and over 5x in GPU memory reduction.
| A simple block contrastive loss approximation technique to efficiently contrast all negative samples, in linear cluster time, at graph level |
Due to depth ambiguities and occlusions, lifting 2D poses to 3D is a highly ill-posed problem. Well-calibrated distributions of possible poses can make these ambiguities explicit and preserve the resulting uncertainty for downstream tasks. This study shows that previous attempts, which account for these ambiguities via multiple hypotheses generation, produce miscalibrated distributions. We identify that miscalibration can be attributed to the use of sample-based metrics such as $\operatorname{minMPJPE}$. In a series of simulations, we show that minimizing $\operatorname{minMPJPE}$, as commonly done, should converge to the correct mean prediction. However, it fails to correctly capture the uncertainty, thus resulting in a miscalibrated distribution. To mitigate this problem, we propose an accurate and well-calibrated model called Conditional Graph Normalizing Flow (cGNFs). Our model is structured such that a single cGNF can estimate both conditional and marginal densities within the same model - effectively solving a zero-shot density estimation problem. We evaluate cGNF on the Human 3.6M dataset and show that cGNF provides a well-calibrated distribution estimate while being close to state-of-the-art in terms of overall $\operatorname{minMPJPE}$. Furthermore, cGNF outperforms previous methods on occluded joints while it remains well-calibrated. | Pose estimation metrics favor overconfident models; we propose cGNF, a model capable of maximizing likelihood and thus estimating accurate and well-calibrated distributions of 3D poses. |
The average reward criterion is relatively less explored as most existing works in the Reinforcement Learning literature consider the discounted reward criterion. There are few recent works that present on-policy average reward actor-critic algorithms, but average reward off-policy actor-critic is relatively less explored. In this paper, we present both on-policy and off-policy deterministic policy gradient theorems for the average reward performance criterion. Using these theorems, we also present an Average Reward Off-Policy Deep Deterministic Policy Gradient (ARO-DDPG) Algorithm. We show a finite time analysis of the resulting three-timescale stochastic approximation scheme and obtain an $\epsilon$-optimal stationary policy with a sample complexity of $\Omega(\epsilon^{-2.5})$. We compare the average reward performance of our proposed algorithm and observe better empirical performance compared to state-of-the-art on-policy average reward actor-critic algorithms over MuJoCo based environments. | This paper proposes actor critic algorithm with deterministic policy for the average reward criterion |
Self-supervised learning (SSL) for graph neural networks (GNNs) has attracted increasing attention from the graph machine learning community in recent years, owing to its capability to learn performant node embeddings without costly label information. One weakness of conventional SSL frameworks for GNNs is that they learn through a single philosophy, such as mutual information maximization or generative reconstruction. When applied to various downstream tasks, these frameworks rarely perform equally well for every task, because one philosophy may not span the extensive knowledge required for all tasks. To enhance the task generalization across tasks, as an important first step forward in exploring fundamental graph models, we introduce PARETOGNN, a multi-task SSL framework for node representation learning over graphs. Specifically, PARETOGNN is self-supervised by manifold pretext tasks observing multiple philosophies. To reconcile different philosophies, we explore a multiple-gradient descent algorithm, such that PARETOGNN actively learns from every pretext task while minimizing potential conflicts. We conduct comprehensive experiments over four downstream tasks (i.e., node classification, node clustering, link prediction, and partition prediction), and our proposal achieves the best overall performance across tasks on 11 widely adopted benchmark datasets. Besides, we observe that learning from multiple philosophies enhances not only the task generalization but also the single task performances, demonstrating that PARETOGNN achieves better task generalization via the disjoint yet complementary knowledge learned from different philosophies. Our code is publicly available at https://github.com/jumxglhf/ParetoGNN. | We present ParetoGNN, a novel multi-task self-supervised learning framework for graph neural networks, that enhances the task generalization across various downstream tasks and datasets. |
Despite the great advances in the machine learning field over the past decade, deep learning algorithms are often vulnerable to data corruption in real-world environments. We propose a simple yet efficient data augmentation method named Exponential Smoothing on Perturbations (ESP) that imposes perturbations on training data to enhance a model’s robustness to unforeseen data corruptions. With the perturbation on the input side, the target label of a sample is smoothed with an exponentially decaying confidence level with respect to the size of the perturbation. ESP enforces a contour-like decision boundary that smoothly encompasses the region around inter-class samples. We theoretically show that perturbations in input space can encourage a model to find a flat minimum on the parameter space, which makes a model robust to domain shifts. In the extensive evaluation on common corruption benchmarks including MNIST-C, CIFAR-10/100-C, and Tiny-ImageNet-C, our method improves the robustness of a model both as a standalone method and in conjunction with the previous state-of-the-art augmentation-based methods. ESP is a model-agnostic algorithm in the sense that it is neither model-specific nor data-specific. | A high-level data augmentation method to increase model robustness against unforeseen data corruptions. |
Inherent heterogeneity of local data distributions, which causes inefficient model learning and significant degradation of model performance, has been a key challenge in Federated Learning (FL). So far, plenty of efforts have focused on addressing data heterogeneity by relying on a hypothetical clustering structure or a consistent information sharing mechanism. However, because of the diversity of the real-world local data, these assumptions may be largely violated. In this work, we argue that information sharing is mostly fragmented in the federated network in reality. More specifically, the distribution overlaps are not consistent but scattered in local clients. In this work, we propose the concept ``Precision Collaboration'' which refers to learning from the informative overlaps precisely while avoiding the potential negative transfer induced by others. In particular, we propose to infer the local data manifolds and estimate the exact local data density simultaneously. The learned manifold aims to precisely identify the overlaps from other clients, and the estimated likelihood allows to generate samples from the manifold in an optimal sampling density. Experiments show that our proposed PCFL significantly overcomes baselines on benchmarks and a real-world clinical scenario. | This paper investigates a precision collaboration mechanism for federated learning. |
Odor perception in mammals is triggered by interactions between volatile organic compounds and a subset of hundreds of proteins called olfactory receptors (ORs). Molecules activate these receptors in a complex combinatorial coding allowing mammals to discriminate a vast number of chemical stimuli. Recently, ORs have gained attention as new therapeutic targets following the discovery of their involvement in other physiological processes and diseases. To date, predicting molecule-induced activation for ORs is highly challenging since $43\%$ of ORs have no identified active compound. In this work, we combine [CLS] token from protBERT with a molecular graph and propose a tailored GNN architecture incorporating inductive biases from the protein-molecule binding. We abstract the biological process of protein-molecule activation as the injection of a molecule into a protein-specific environment. On a newly gathered dataset of $46$ $700$ OR-molecule pairs, this model outperforms state-of-the-art models on drug-target interaction prediction as well as standard GNN baselines. Moreover, by incorporating non-bonded interactions the model is able to work with mixtures of compounds. Finally, our predictions reveal a similar activation pattern for molecules within a given odor family, which is in agreement with the theory of combinatorial coding in olfaction. | We leverage recent advances in protein representation learning and graph neural networks to predict olfactory receptor-molecule binding. |
We address the problem of safe reinforcement learning from pixel observations. Inherent challenges in such settings are (1) a trade-off between reward optimization and adhering to safety constraints, (2) partial observability, and (3) high-dimensional observations. We formalize the problem in a constrained, partially observable Markov decision process framework, where an agent obtains distinct reward and safety signals. To address the curse of dimensionality, we employ a novel safety critic using the stochastic latent actor-critic (SLAC) approach. The latent variable model predicts rewards and safety violations, and we use the safety critic to train safe policies. Using well-known benchmark environments, we demonstrate competitive performance over existing approaches regarding computational requirements, final reward return, and satisfying the safety constraints. | This paper proposes Safe SLAC, a safety-constrained RL approach for partially observable settings, which uses a stochastic latent variable model combined with a safety critic. |
Recently, the Successor Features and Generalized Policy Improvement (SF&GPI) framework has been proposed as a method for learning, composing and transferring predictive knowledge and behavior. SF&GPI works by having an agent learn predictive representations (SFs) that can be combined for transfer to new tasks with GPI. However, to be effective this approach requires state features that are useful to predict, and these state-features are typically hand-designed. In this work, we present a novel neural network architecture, “Modular Successor Feature Approximators” (MSFA), where modules both discover what is useful to predict, and learn their own predictive representations. We show that MSFA is able to better generalize compared to baseline architectures for learning SFs and a modular network that discovers factored state representations.
| A modular neural network for discovering, composing, and transferring predictive knowledge and behavior via Successor Features & Generalized Policy Improvement. |
We consider monotone variational inequality (VI) problems in multi-GPU settings where multiple processors/workers/clients have access to local stochastic dual vectors. This setting includes a broad range of important problems from distributed convex minimization to min-max and games. Extra-gradient, which is a de facto algorithm for monotone VI problems, has not been designed to be communication-efficient. To this end, we propose a quantized generalized extra-gradient (Q-GenX), which is an unbiased and adaptive compression method tailored to solve VIs. We provide an adaptive step-size rule, which adapts to the respective noise profiles at hand and achieve a fast rate of ${\cal O}(1/T)$ under relative noise, and an order-optimal ${\cal O}(1/\sqrt{T})$ under absolute noise and show distributed training accelerates convergence. Finally, we validate our theoretical results by providing real-world experiments and training generative adversarial networks on multiple GPUs.
| We propose quantized generalized extra-gradient, which is an unbiased and adaptive compression method tailored to a generic unifying framework for solving variational inequality problems. |
The success of machine learning relies heavily on massive amounts of data, which are usually generated and stored across a range of diverse and distributed data sources. Decentralized learning has thus been advocated and widely deployed to make efficient use of distributed datasets, with an extensive focus on supervised learning (SL) problems. Unfortunately, the majority of real-world data are unlabeled and can be highly heterogeneous across sources. In this work, we carefully study decentralized learning with unlabeled data through the lens of self-supervised learning (SSL), specifically contrastive visual representation learning. We study the effectiveness of a range of contrastive learning algorithms under a decentralized learning setting, on relatively large-scale datasets including ImageNet-100, MS-COCO, and a new real-world robotic warehouse dataset. Our experiments show that the decentralized SSL (Dec-SSL) approach is robust to the heterogeneity of decentralized datasets, and learns useful representation for object classification, detection, and segmentation tasks, even when combined with the simple and standard decentralized learning algorithm of Federated Averaging (FedAvg). This robustness makes it possible to significantly reduce communication and to reduce the participation ratio of data sources with only minimal drops in performance. Interestingly, using the same amount of data, the representation learned by Dec-SSL can not only perform on par with that learned by centralized SSL which requires communication and excessive data storage costs, but also sometimes outperform representations extracted from decentralized SL which requires extra knowledge about the data labels. Finally, we provide theoretical insights into understanding why data heterogeneity is less of a concern for Dec-SSL objectives, and introduce feature alignment and clustering techniques to develop a new Dec-SSL algorithm that further improves the performance, in the face of highly non-IID data. Our study presents positive evidence to embrace unlabeled data in decentralized learning, and we hope to provide new insights into whether and why decentralized SSL is effective and/or even advantageous. | We study decentralized learning with non-IID unlabeled data, and try to understand the robustness and communication efficiency of decentralized self-supervised learning, through extensive experiments and theoretical analysis. |
Recent years have seen progress beyond domain-specific sound separation for speech or music towards universal sound separation for arbitrary sounds. Prior work on universal sound separation has investigated separating a target sound out of an audio mixture given a text query. Such text-queried sound separation systems provide a natural and scalable interface for specifying arbitrary target sounds. However, supervised text-queried sound separation systems require costly labeled audio-text pairs for training. Moreover, the audio provided in existing datasets is often recorded in a controlled environment, causing a considerable generalization gap to noisy audio in the wild. In this work, we aim to approach text-queried universal sound separation by using only unlabeled data. We propose to leverage the visual modality as a bridge to learn the desired audio-textual correspondence. The proposed CLIPSep model first encodes the input query into a query vector using the contrastive language-image pretraining (CLIP) model, and the query vector is then used to condition an audio separation model to separate out the target sound. While the model is trained on image-audio pairs extracted from unlabeled videos, at test time we can instead query the model with text inputs in a zero-shot setting, thanks to the joint language-image embedding learned by the CLIP model. Further, videos in the wild often contain off-screen sounds and background noise that may hinder the model from learning the desired audio-textual correspondence. To address this problem, we further propose an approach called noise invariant training for training a query-based sound separation model on noisy data. Experimental results show that the proposed models successfully learn text-queried universal sound separation using only noisy unlabeled videos, even achieving competitive performance against a supervised model in some settings. | A new method the leverages the pretrained CLIP model and noise invariant training for learning text-queried sound separation with only noisy unlabeled videos |
We develop an interior-point approach to solve constrained variational inequality (cVI) problems. Inspired by the efficacy of the alternating direction method of multipliers (ADMM) method in the single-objective context, we generalize ADMM to derive a first-order method for cVIs, that we refer to as ADMM-based interior-point method for constrained VIs (ACVI). We provide convergence guarantees for ACVI in two general classes of problems: (i) when the operator is $\xi$-monotone, and (ii) when it is monotone, some constraints are active and the game is not purely rotational. When the operator is in addition L-Lipschitz for the latter case, we match known lower bounds on rates for the gap function of $\mathcal{O}(1/\sqrt{K})$ and $\mathcal{O}(1/K)$ for the last and average iterate, respectively. To the best of our knowledge, this is the first presentation of a first-order interior-point method for the general cVI problem that has a global convergence guarantee. Moreover, unlike previous work in this setting, ACVI provides a means to solve cVIs when the constraints are nontrivial. Empirical analyses demonstrate clear advantages of ACVI over common first-order methods. In particular, (i) cyclical behavior is notably reduced as our methods approach the solution from the analytic center, and (ii) unlike projection-based methods that zigzag when near a constraint, ACVI efficiently handles the constraints. | We derive a first-order method for solving constrained variational inequality problem when given general constraints, by combining interior-point methods and ADMM. |
Uncertainty-aware modeling has emerged as a key component in modern machine learning frameworks. The de-facto standard approach adopts heteroscedastic Gaussian distributions and minimizes the negative log-likelihood (NLL) under observed data. However, optimizing this objective turns out to be surprisingly intricate, and the current state-of-the-art reports several instabilities. This work breaks down the optimization problem, initially focusing on non-contextual settings where convergence can be analyzed analytically. We show that (1) in this learning scheme, the eigenvalues of the predictive covariance define stability in learning, and (2) coupling of gradients and predictions build up errors in both mean and covariance if either is poorly approximated. Building on these insights, we propose Trustable, a novel optimizer that overcomes instabilities methodically by combining systematic update restrictions in the form of trust regions with structured, tractable natural gradients. We demonstrate in several challenging experiments that Trustable outperforms current optimizers in regression with neural networks in terms of the NLL, MSE, and further performance metrics. Unlike other optimizers, Trustable yields an improved and more stable fit and can also be applied to multivariate outputs with full covariance matrices. | We analyze the instability of Gaussian likelihood optimization and propose a gradient-based optimizer demonstrating less volatile optimization especially for contextual, multivariate target distributions with full covariances. |
Multi-view projection methods have demonstrated promising performance on 3D understanding tasks like 3D classification and segmentation. However, it remains unclear how to combine such multi-view methods with the widely available 3D point clouds. Previous methods use unlearned heuristics to combine features at the point level. To this end, we introduce the concept of the multi-view point cloud (Voint cloud), representing each 3D point as a set of features extracted from several view-points. This novel 3D Voint cloud representation combines the compactness of 3D point cloud representation with the natural view-awareness of multi-view representation. Naturally, we can equip this new representation with convolutional and pooling operations. We deploy a Voint neural network (VointNet) to learn representations in the Voint space. Our novel representation achieves state-of-the-art performance on 3D classification, shape retrieval, and robust 3D part segmentation on standard benchmarks ( ScanObjectNN, ShapeNet Core55, and ShapeNet Parts). Further analysis shows that VointNet improves the robustness to occlusion compared to other methods. | We propose voint cloud, a novel 3D data structure, that combines multi-view and point clouds for robust 3D understanding tasks. |
Neural networks can be drastically shrunk in size by removing redundant parameters. While crucial for the deployment on resource-constraint hardware, oftentimes, compression comes with a severe drop in accuracy and lack of adversarial robustness. Despite recent advances, counteracting both aspects has only succeeded for moderate compression rates so far. We propose a novel method, HARP, that copes with aggressive pruning significantly better than prior work. For this, we consider the network holistically. We learn a global compression strategy that optimizes how many parameters (compression rate) and which parameters (scoring connections) to prune specific to each layer individually. Our method fine-tunes an existing model with dynamic regularization, that follows a step-wise incremental function balancing the different objectives. It starts by favoring robustness before shifting focus on reaching the target compression rate and only then handles the objectives equally. The learned compression strategies allow us to maintain the pre-trained model’s natural accuracy and its adversarial robustness for a reduction by 99% of the network’s original size. Moreover, we observe a crucial influence of non-uniform compression across layers. The implementation of HARP is publicly available at https://intellisec.de/research/harp. | We propose HARP that realizes the adversarially robust pruning in a holistic way and yields an outstanding capability at aggressive compression. |
Like many other machine learning applications, neural machine translation (NMT) benefits from over-parameterized deep neural models. However, these models have been observed to be brittle: NMT model predictions are sensitive to small input changes and can show significant variation across re-training or incremental model updates. This work studies a frequently used method in NMT, pseudo-label training (PLT), which is common to the related techniques of forward-translation (or self-training) and sequence-level knowledge distillation. While the effect of PLT on quality is well-documented, we highlight a lesser-known effect: PLT can enhance a model's stability to model updates and input perturbations, a set of properties we call \textit{model inertia}. We study inertia effects under different training settings and we identify distribution simplification as a mechanism behind the observed results. | pseudo-label training improves model stablity to updates and input perturbations |
Randomized experiments (a.k.a. A/B tests) are a powerful tool for estimating treatment effects, to inform decisions making in business, healthcare and other applications. In many problems, the treatment has a lasting effect that evolves over time. A limitation with randomized experiments is that they do not easily extend to measure long-term effects, since running long experiments is time-consuming and expensive. In this paper, we take a reinforcement learning (RL) approach that estimates the average reward in a Markov process. Motivated by real-world scenarios where the observed state transition is nonstationary, we develop a new algorithm for a class of nonstationary problems, and demonstrate promising results in two synthetic datasets and one online store dataset. | We propose a reinforcement learning approach to estimating the long-term treatment effect from short-term data. |
Bias is a common problem inherent in recommender systems, which is entangled with users' preferences and poses a great challenge to unbiased learning. For debiasing tasks, the doubly robust (DR) method and its variants show superior performance due to the double robustness property, that is, DR is unbiased when either imputed errors or learned propensities are accurate.
However, our theoretical analysis reveals that DR usually has a large variance. Meanwhile, DR would suffer unexpectedly large bias and poor generalization caused by inaccurate imputed errors and learned propensities, which usually occur in practice. In this paper, we propose a principled approach that can effectively reduce the bias and variance simultaneously for existing DR approaches when the error imputation model is misspecified. In addition, we further propose a novel semi-parametric collaborative learning approach that decomposes imputed errors into parametric and nonparametric parts and updates them collaboratively, resulting in more accurate predictions. Both theoretical analysis and experiments demonstrate the superiority of the proposed methods compared with existing debiasing methods. | This paper proposes a principled approach that can effectively reduce the bias and variance simultaneously compared to existing DR estimators for debiased recommendations. |
The \emph{stability} of learning algorithms to changes in the training sample has been actively studied as a powerful proxy for reasoning about generalization. Recently, exponential generalization and excess risk bounds with near-optimal rates have been obtained under the stringent and distribution-free notion of uniform stability~\citep{bousquet2020sharper,klochkov2021stability}. In the meanwhile, under the notion of $L_q$-stability, which is weaker and distribution dependent, exponential generalization bounds are also available yet so far only with sub-optimal rates. Therefore, a fundamental question we would like to address in this paper is whether it is possible to derive near-optimal exponential generalization bounds for $L_q$-stable learning algorithms. As the core contribution of the present work, we give an affirmative answer to this question by developing strict analogues of the near-optimal generalization and risk bounds of uniformly stable algorithms for $L_q$-stable algorithms. Further, we demonstrate the power of our improved $L_q$-stability and generalization theory by applying it to derive strong sparse excess risk bounds, under mild conditions, for computationally tractable sparsity estimation algorithms such as Iterative Hard Thresholding (IHT). | We presented a set of sharper and near-optimal exponential generalization bounds for $L_q$-stable learning algorithms |
We consider the problem of learning control policies in stochastic systems which guarantee that the system stabilizes within some specified stabilization region with probability 1. Our approach is based on the novel notion of stabilizing ranking supermartingales (sRSMs) that we introduce in this work. Our sRSMs overcome the limitation of methods proposed in previous works whose applicability is restricted to systems in which the stabilizing region cannot be left once entered under any control policy. We present a learning procedure that learns a control policy together with an sRSM that formally certifies probability 1 stability, both learned as neural networks. Our experimental evaluation shows that our learning procedure can successfully learn provably stabilizing policies in practice. | We learn policies and certificates for proving region stabilization in control systems |
Recent studies have revealed the intriguing few-shot learning ability of pretrained language models (PLMs): They can quickly adapt to a new task when fine-tuned on a small amount of labeled data formulated as prompts, without requiring abundant task-specific annotations. Despite their promising performance, most existing few-shot approaches that only learn from the small training set still underperform fully supervised training by nontrivial margins. In this work, we study few-shot learning with PLMs from a different perspective: We first tune an autoregressive PLM on the few-shot samples and then use it as a generator to synthesize a large amount of novel training samples which augment the original training set. To encourage the generator to produce label-discriminative samples, we train it via weighted maximum likelihood where the weight of each token is automatically adjusted based on a discriminative meta-learning objective. A classification PLM can then be fine-tuned on both the few-shot and the synthetic samples with regularization for better generalization and stability. Our approach FewGen achieves an overall better result across seven classification tasks of the GLUE benchmark than existing few-shot learning methods. | We study how to effectively tune language models to generate training samples for few-shot learning. |
Continual learning (CL) has two main objectives: preventing catastrophic forgetting (CF) and encouraging knowledge transfer (KT) across tasks. The existing literature mainly tries to overcome CF. Although some papers have focused on both CF and KT, they may still suffer from CF because of their ineffective handling of previous tasks and/or poor task similarity detection mechanisms to achieve KT. This work presents a new CL method that addresses the above issues. First, it overcomes CF by isolating the knowledge of each task via a learned mask that indicates a sub-network. Second, it proposes a novel technique to compute how important each mask is to the new task, which indicates how the new task is similar to an underlying old task. Similar tasks can share the same mask/subnetwork for KT, while dissimilar tasks use different masks/sub-networks for CF prevention. Comprehensive experiments have been conducted using a range of NLP problems, including classification, generation, and extraction to show that the proposed method consistently outperforms prior state-of-the-art baselines. | A continual learning method based on sub-networks and task similarity is proposed and evaluated on NLP classification, generation, and extraction problems. |
We consider sketching algorithms which first compress data by multiplication with a random sketch matrix, and then apply the sketch to quickly solve an optimization problem, e.g., low-rank approximation and regression. In the learning-based sketching paradigm proposed by Indyk et al., the sketch matrix is found by choosing a random sparse matrix, e.g., CountSketch, and then the values of its non-zero entries are updated by running gradient descent on a training data set. Despite the growing body of work on this paradigm, a noticeable omission is that the locations of the non-zero entries of previous algorithms were fixed, and only their values were learned.
In this work, we propose the first learning-based algorithms that also optimize the locations of the non-zero entries. Our first proposed algorithm is based on a greedy algorithm. However, one drawback of the greedy algorithm is its slower training time. We fix this issue and propose approaches for learning a sketching matrix for both low-rank approximation and Hessian approximation for second-order optimization. The latter is helpful for a range of constrained optimization problems, such as LASSO and matrix estimation with a nuclear norm constraint. Both approaches achieve good accuracy with a fast running time. Moreover, our experiments suggest that our algorithm can still reduce the error significantly even if we only have a very limited number of training matrices. | We propose the first learning-based algorithms that also optimize the locations of the non-zero entries of CountSketch matrix. |
Multi-Agent Reinforcement Learning currently focuses on implementations where all data and training can be centralized to one machine. But what if local agents are split across multiple tasks, and need to keep data private between each? We develop the first application of Personalized Federated Hypernetworks (PFH) to Reinforcement Learning (RL). We then present a novel application of PFH to few-shot transfer, and demonstrate significant initial increases in learning. PFH has never been demonstrated beyond supervised learning benchmarks, so we apply PFH to an important domain: RL price-setting for energy demand response. We consider a general case across where agents are split across multiple microgrids, wherein energy consumption data must be kept private within each microgrid. Together, our work explores how the fields of personalized federated learning and RL can come together to make learning efficient across multiple tasks while keeping data secure. | We use hypernetworks to aggregate learning across multiple reinforcement learning agents in a microgrid energy demand response setting while preserving privacy. |
Computing Nash equilibrium policies is a central problem in multi-agent reinforcement learning that has received extensive attention both in theory and in practice. However, in light of computational intractability barriers in general-sum games, provable guarantees have been thus far either limited to fully competitive or cooperative scenarios or impose strong assumptions that are difficult to meet in most practical applications.
In this work, we depart from those prior results by investigating infinite-horizon \emph{adversarial team Markov games}, a natural and well-motivated class of games in which a team of identically-interested players---in the absence of any explicit coordination or communication---is competing against an adversarial player. This setting allows for a unifying treatment of zero-sum Markov games and Markov potential games, and serves as a step to model more realistic strategic interactions that feature both competing and cooperative interests. Our main contribution is the first algorithm for computing stationary $\epsilon$-approximate Nash equilibria in adversarial team Markov games with computational complexity that is polynomial in all the natural parameters of the game, as well as $1/\epsilon$.
The proposed algorithm is based on performing independent policy gradient steps for each player in the team, in tandem with best responses from the side of the adversary; in turn, the policy for the adversary is then obtained by solving a carefully constructed linear program. Our analysis leverages non-standard techniques to establish the KKT optimality conditions for a nonlinear program with nonconvex constraints, thereby leading to a natural interpretation of the induced Lagrange multipliers. | Nash equlibrium can be computed effieciently in Markov games where a single player competes against multiple agents who share a common interest. |
The social behavior change in a population has long been studied as an essential component of multi-agent learning. The learning of behavioral change not only involves reinforcement learning (RL), but also be measured against the general population with mutual information (MI). The combination of RL and MI led us to derive MI optimizations from policy gradient. With MI as multi-agent's optimization objective, we discover that the dual properties of MI can result in distinctly different population behaviors. From MI maximization that maximizes the stability of a population to MI minimization that enables specialization among the agents, the dual of MI creates a significant change in a population's behavioral properties. In this paper, we propose a minimax formulation of MI (M\&M) that enables agents specialization with stable regularization. Empirically we evaluated M\&M against the prior SOTA MARL framework, and analyze the social behavior change in performance, diversity, and the stability of their social graphs. | The social behavioral change in population learning is impacted by the dual properties of mutual information. |
Recently, generalization on out-of-distribution (OOD) data with correlation shift has attracted great attentions. The correlation shift is caused by the spurious attributes that correlate to the class label, as the correlation between them may vary in training and test data. For such a problem, we show that given the class label, the models that are conditionally independent of spurious attributes are OOD generalizable. Based on this, a metric Conditional Spurious Variation (CSV) which controls the OOD generalization error, is proposed to measure such conditional independence. To improve the OOD generalization, we regularize the training process with the proposed CSV. Under mild assumptions, our training objective can be formulated as a nonconvex-concave mini-max problem. An algorithm with a provable convergence rate is proposed to solve the problem. Extensive empirical results verify our algorithm's efficacy in improving OOD generalization. | This paper proposes an algorithm to make the model to generalize on data with spurious correlation, the method can be implemented without information on spurious feature. |
Large language models (LMs) have been shown to memorize parts of their training data, and when prompted appropriately, they will emit the memorized training data verbatim. This is undesirable because memorization violates privacy (exposing user data), degrades utility (repeated easy-to-memorize text is often low quality), and hurts fairness (some texts are memorized over others).
We describe three log-linear relationships that quantify the degree to which LMs emit memorized training data. Memorization significantly grows as we increase (1) the capacity of a model, (2) the number of times an example has been duplicated, and (3) the number of tokens of context used to prompt the model. Surprisingly, we find the situation becomes complicated when generalizing these results across model families. On the whole, we find that memorization in LMs is more prevalent than previously believed and will likely get worse as models continues to scale, at least without active mitigations. | Model size, duplication, and context length all impact how easy it is to extract training data from large language models. |
Understanding the input-output mapping relationship in the \emph{entire input space} contributes a novel perspective to a comprehensive understanding of deep neural networks. In this paper, we focus on binary neural classifiers and propose to first uncover the histogram about the number of inputs that are mapped to certain output values and then scrutinize the representative inputs from a certain output range of interest, such as the positive-logit region that corresponds to one of the classes. A straightforward solution is uniform sampling (or exhaustive enumeration) in the entire input space but when the inputs are high dimensional, it can take almost forever to converge. We connect the output histogram to the \emph{density of states} in physics by making an analogy between the energy of a system and the neural network output. Inspired by the Wang-Landau algorithm designed for sampling the density of states, we propose an efficient sampler that is driven to explore the under-explored output values through a gradient-based proposal. Compared with the random proposal in Wang-Landau algorithm, our gradient-based proposal converges faster as it can propose the inputs corresponding to the under-explored output values. Extensive experiments have verified the accuracy of the histogram generated by our sampler and also demonstrated interesting findings. For example, the models map many human unrecognizable images to very negative logit values. These properties of a neural model are revealed for the first time through our sampled statistics. We believe that our approach opens a new gate for neural model evaluation and shall be further explored in future works.
| We draw the connection between energy in physics and output of neural networks and propose an efficient sampler for better understanding of the input-output mapping relationship of the (binary) neural classifiers. |
Graph augmentation is the key component to reveal instance-discriminative features of a graph as its rationale in graph contrastive learning (GCL).
And existing rationale-aware augmentation mechanisms in GCL frameworks roughly fall into two categories and suffer from inherent limitations: (1) non-heuristic methods with the guidance of domain knowledge to preserve salient features, which require expensive expertise and lacks generality, or (2) heuristic augmentations with a co-trained auxiliary model to identify crucial substructures, which face not only the dilemma between system complexity and transformation diversity, but also the instability stemming from the co-training of two separated sub-models.
Inspired by recent studies on transformers, we propose $\underline{S}$elf-attentive $\underline{R}$ationale guided $\underline{G}$raph $\underline{C}$ontrastive $\underline{L}$earning (SR-GCL), which integrates rationale finder and encoder together, leverages the self-attention values in transformer module as a natural guidance to delineate semantically informative substructures from both node- and edge-wise views, and contrasts on rationale-aware augmented pairs.
On real world biochemistry datasets, visualization results verify the effectiveness of self-attentive rationalization and the performance on downstream tasks demonstrates the state-of-the-art performance of SR-GCL for graph model pre-training. | Graph contrastive learning framework with self-attentive rationalization |
Self-training (ST), or pseudo-labeling has sparked significant interest in the automatic speech recognition (ASR) community recently because of its success in harnessing unlabeled data. Unlike prior semi-supervised learning approaches that relied on iteratively regenerating pseudo-labels (PLs) from a trained model and using them to train a new model, recent state-of-the-art methods perform `continuous training' where PLs are generated using a very recent version of the model being trained. Nevertheless, these approaches still rely on bootstrapping the ST using an initial supervised learning phase where the model is trained on labeled data alone. We believe this has the potential for over-fitting to the labeled dataset in low resource settings and that ST from the start of training should reduce over-fitting. In this paper we show how we can do this by dynamically controlling the evolution of PLs during the training process in ASR. To the best of our knowledge, this is the first study that shows the feasibility of generating PLs from the very start of the training. We are able to achieve this using two techniques that avoid instabilities which lead to degenerate models that do not generalize. Firstly, we control the evolution of PLs through a curriculum that uses the online changes in PLs to control the membership of the cache of PLs and improve generalization. Secondly, we find that by sampling transcriptions from the predictive distribution, rather than only using the best transcription, we can stabilize training further. With these techniques, our ST models match prior works without an external language model. | We show how to perform continuous self-training right from the start without any supervised pre-training. |
Existing machine learning methods for causal inference usually estimate quantities expressed via the mean of potential outcomes (e.g., average treatment effect). However, such quantities do not capture the full information about the distribution of potential outcomes. In this work, we estimate the density of potential outcomes after interventions from observational data. For this, we propose a novel, fully-parametric deep learning method called Interventional Normalizing Flows. Specifically, we combine two normalizing flows, namely (i) a teacher flow for estimating nuisance parameters and (ii) a student flow for a parametric estimation of the density of potential outcomes. We further develop a tractable optimization objective via a one-step bias correction for an efficient and doubly robust estimation of the student flow parameters. As a result our Interventional Normalizing Flows offer a properly normalized density estimator. Across various experiments, we demonstrate that our Interventional Normalizing Flows are expressive and highly effective, and scale well with both sample size and high-dimensional confounding. To the best of our knowledge, our Interventional Normalizing Flows are the first fully-parametric, deep learning method for density estimation of potential outcomes. | We propose a novel, fully-parametric deep learning method for estimating densities of potential outcomes, called Interventional Normalizing Flows |
Many real-world reinforcement learning tasks require control of complex dynamical systems that involve both costly data acquisition processes and large state spaces. In cases where the expensive transition dynamics can be readily evaluated at specified states (e.g., via a simulator), agents can operate in what is often referred to as planning with a \emph{generative model}. We propose the AE-LSVI algorithm for best policy identification, a novel variant of the kernelized least-squares value iteration (LSVI) algorithm that combines optimism with pessimism for active exploration (AE). AE-LSVI provably identifies a near-optimal policy \emph{uniformly} over an entire state space and achieves polynomial sample complexity guarantees that are independent of the number of states. When specialized to the recently introduced offline contextual Bayesian optimization setting, our algorithm achieves improved sample complexity bounds. Experimentally, we demonstrate that AE-LSVI outperforms other RL algorithms in a variety of environments when robustness to the initial state is required. | We propose a novel kernelized LSVI algorithm for active reinforcement learning which provably identifies a near-optimal policy uniformly over the entire state space. |
The objective of drug discovery is to find novel compounds with desirable chemical properties. Generative models have been utilized to sample molecules at the intersection of multiple property constraints. In this paper we pose molecular design as a language modeling problem where the model implicitly learns the vocabulary and composition of valid molecules, hence it is able to generate new molecules of interest. We present SmilesFormer, a Transformer-based model which is able to encode molecules, molecule fragments, and fragment compositions as latent variables, which are in turn decoded to stochastically generate novel molecules. This is achieved by fragmenting the molecules into smaller combinatorial groups, then learning the mapping between the input fragments and valid SMILES sequences.
The model is able to optimize molecular properties through a stochastic latent space traversal technique. This technique systematically searches the encoded latent space to find latent vectors that are able to produce molecules to meet the multi-property objective. The model was validated through various de novo molecular design tasks, achieving state-of-the-art performances when compared to previous methods. Furthermore, we used the proposed method to demonstrate a drug rediscovery pipeline for Donepezil, a known Acetylcholinesterase Inhibitor. | We developed a transformer-based language model for SMILES strings, able to generate and efficiently optimize molecules for drug discovery. |
Diffusion Probabilistic Models (DPMs) achieve impressive performance in visual generative tasks recently. However, the success of DPMs heavily relies on large amounts of data and optimization steps, which limits the application of DPMs to small datasets and limited computational resources. In this paper, we investigate transfer learning in DPMs to leverage the DPMs pretrained on large-scale datasets for generation with limited data. Firstly, we show that previous methods like training from scratch or determining the transferable parts is not suitable for the DPM due to its U-Net based denoising architecture with the external denoising timestep input. To address it, we present a condition-based tuning approach to take full advantages of existing pretrained models. Concretely, we obtain the semantic embeddings of condition images by the pretrained CLIP model, and then inject these semantic informations to the pretrained DPM via a ''Attention-NonLinear'' (ANL) module. The adaptation to a new task can be achieved by only tuning the ANL module inserted into the pretrained DPM hierarchically. To further enhance the diversity of generated images, we introduce a masked sampling strategy based on the condition mechanism. Extensive experiments validate the effectiveness and efficiency of our proposed tuning approach in generative task transfer and data augmentation for semi-supervised learning. | We propose a new tuning approach for transferring pretrained diffusion probabilistic models to new tasks with limited data and training resources. |
Denoising diffusion models have recently marked a milestone in high-quality image generation. One may thus wonder if they are suitable for neural image compression. This paper outlines an end-to-end optimized image compression framework based on a conditional diffusion model, drawing on the transform-coding paradigm. Besides the latent variables inherent to the diffusion process, this paper introduces an additional discrete "content" latent variable to condition the denoising process on. This variable is equipped with a hierarchical prior for entropy coding. The remaining "texture" latent variables characterizing the diffusion process are synthesized (either stochastically or deterministically) at decoding time. We furthermore show that the performance can be tuned toward perceptual metrics of interest. Our extensive experiments involving five datasets and 16 image perceptual quality assessment metrics show that our approach not only compares favorably in terms of rate and perceptual distortion tradeoffs but also shows robust performance under all metrics while other baselines show less consistent behavior. | We show that hybridizing compressive VAEs with denoising diffusion models leads to strong performance in perceptual image compression.. |
Proteins are complex molecules responsible for different functions in the human body. Enhancing the functionality of a protein and/or cellular fitness can significantly impact various industries. However, their optimization remains challenging, and sequences generated by data-driven methods often fail in wet lab experiments. This study investigates the limitations of existing model-based sequence design methods and presents a novel optimization framework that can efficiently traverse the latent representation space instead of the protein sequence space. Our framework generates proteins with higher functionality and cellular fitness by modeling the sequence design task as a Markov decision process and applying model-based reinforcement learning. We discuss the results in a comprehensive evaluation of two distinct proteins, GPF and His3, along with the predicted structure of optimized sequences using deep learning-based structure prediction. | This study investigates why many model-based biological sequence design methods produce results that empirically fail and proposes a novel optimization process that can efficiently traverse a latent representation space instead of the sequence space. |
Many important problems involving molecular property prediction from 3D structures have limited data, posing a generalization challenge for neural networks. In this paper, we describe a pre-training technique based on denoising that achieves a new state-of-the-art in molecular property prediction by utilizing large datasets of 3D molecular structures at equilibrium to learn meaningful representations for downstream tasks. Relying on the well-known link between denoising autoencoders and score-matching, we show that the denoising objective corresponds to learning a molecular force field -- arising from approximating the Boltzmann distribution with a mixture of Gaussians -- directly from equilibrium structures. Our experiments demonstrate that using this pre-training objective significantly improves performance on multiple benchmarks, achieving a new state-of-the-art on the majority of targets in the widely used QM9 dataset. Our analysis then provides practical insights into the effects of different factors -- dataset sizes, model size and architecture, and the choice of upstream and downstream datasets -- on pre-training. | We describe a technique for pre-training models for molecular property prediction from 3D structures based on denoising and show that it achieves SOTA results for various tasks. |
Photorealistic style transfer aims to transfer the artistic style of an image onto an input image or video while keeping photorealism. In this paper, we think it's the summary statistics matching scheme in existing algorithms that leads to unrealistic stylization. To avoid employing the popular Gram loss, we propose a self-supervised style transfer framework, which contains a style removal part and a style restoration part. The style removal network removes the original image styles, and the style restoration network recovers image styles in a supervised manner. Meanwhile, to address the problems in current feature transformation methods, we propose decoupled instance normalization to decompose feature transformation into style whitening and restylization. It works quite well in ColoristaNet and can transfer image styles efficiently while keeping photorealism. To ensure temporal coherency, we also incorporate optical flow methods and ConvLSTM to embed contextual information. Experiments demonstrates that ColoristaNet can achieve better stylization effects when compared with state-of-the-art algorithms. | In this paper, we propose a novel photorealistic video style transfer network called ColoristaNet, which can conduct color style transfer in videos without introducing any painterly spatial distortions and inconsistent flickering artifacts. |
Accurate uncertainty measurement is a key step to building robust and reliable machine learning systems. Conformal prediction is a distribution-free uncertainty quantification algorithm popular for its ease of implementation, statistical coverage guarantees, and versatility for underlying forecasters. However, existing conformal prediction algorithms for time series are limited to single-step prediction without considering the temporal dependency. In this paper we propose a Copula-based Conformal Prediction algorithm for multivariate, multi-step Time Series forecasting, CopulaCPTS. On several synthetic and real-world multivariate time series datasets, we show that CopulaCPTS produces more calibrated and sharp confidence intervals for multi-step prediction tasks than existing techniques. | significantly improve efficiency/sharpness of conformal prediction confidence intervals, for time series forecasting, by modeling dependence of time steps using copulas |
Recently introduced Contrastive Language-Image Pre-Training (CLIP) bridges images and text by embedding them into a joint latent space. This opens the door to ample literature that aims to manipulate an input image by providing a textual explanation. However, due to the discrepancy between image and text embeddings in the joint space, using text embeddings as the optimization target often introduces undesired artifacts in the resulting images. Disentanglement, interpretability, and controllability are also hard to guarantee for manipulation. To alleviate these problems, we propose to define corpus subspaces spanned by prompts to capture specific image characteristics. We introduce CLIP projection-augmentation embedding (PAE) as an optimization target to improve the performance of text-guided image manipulation. Our method is a simple and general paradigm that can be easily computed and adapted, and smoothly incorporated into any CLIP-based latent manipulation algorithm to improve performance. To demonstrate the effectiveness of our method, we conduct several theoretical and empirical system studies. As a case study, we utilize the method for text-guided semantic face editing. We quantitatively and qualitatively demonstrate that PAE facilitates a more disentangled, interpretable, and controllable image manipulation method with state of the art quality and accuracy. | We propose a novel approach to enforce better disentanglement, interpretability and controllability for text-guided image manipulation. |
Adaptive optimizers (e.g., Adam) have achieved tremendous success in deep learning. The key component of the optimizer is the precondition matrix, which provides more gradient information and adjusts the step size of each gradient direction. Intuitively, the closer the precondition matrix approximates the Hessian, the faster convergence and better generalization the optimizer can achieve in terms of iterations. However, this performance improvement is usually accompanied by a huge increase in the amount of computation. In this paper, we propose a new optimizer called AdaDQH to achieve better generalization with acceptable computational overhead. The intuitions are the trade-off of the precondition matrix between computation time and approximation of Hessian, and the auto switch of the precondition matrix from Stochastic Gradient Descent (SGD) to the adaptive optimizer. We evaluate AdaDQH on public datasets of Computer Vision (CV), Natural Language Processing (NLP) and Recommendation Systems (RecSys). The experimental results reveal that, compared to the State-Of-The-Art (SOTA) optimizers, AdaDQH can achieve significantly better or highly competitive performance. Furthermore, we analyze how AdaDQH is able to auto switch from stochastic to adaptive and the actual effects in different scenes. The code is available in the supplemental material. | We propose the AdaDQH optimizer, which can evolve from stochastic to adaptive by auto switch of the precondition matrix and has better performance compared to the state-of-the-art optimizers. |
The success of modern neural models has prompted renewed study of the connection between memorisation and generalisation: such models typically generalise well, despite being able to perfectly fit ("memorise") completely random labels.
To more carefully study this issue, Feldman (2019); Feldman & Zhang (2020) provided a simple metric to quantify the degree of memorisation of a specific training example, and empirically quantified the corresponding memorisation profile of a ResNet model on image classification benchmarks.
While an exciting first glimpse into how real-world models memorise, these studies leave open several questions about memorisation of practical networks.
In particular, how is memorisation affected by increasing model size, and by distilling a large model into a smaller one?
We present a systematic empirical analysis of these questions.
On standard image classification benchmarks, we find that training examples exhibit a diverse set of memorisation trajectories across model sizes, with some samples having increased memorisation under larger models.
Further, we find that distillation tends to inhibit memorisation of the student model, while also improving generalisation.
Finally, we show that computationally tractable measures of memorisation do not capture the properties we identify for memorisation in the sense of Feldman (2019), despite highly correlating to the latter. | Increasing model size may increase memorisation of certain training samples, while distillation inhibits memorisation |
Traditional Dictionary Learning (DL) aims to approximate data vectors as sparse linear combinations of basis elements (atoms) and is widely used in machine learning, computer vision, and signal processing. To extend DL to graphs, Vincent-Cuaz et al. 2021 propose a method, called GDL, which describes the topology of each graph with a pairwise relation matrix (PRM) and compares PRMs via the Gromov-Wasserstein Discrepancy (GWD). However, the lack of robustness often excludes GDL from a variety of real-world applications since GWD is sensitive to the structural noise in graphs. This paper proposes an improved graph dictionary learning algorithm based on a robust Gromov-Wasserstein discrepancy (RGWD) which has theoretically sound properties and an efficient numerical scheme. Based on such a discrepancy, our dictionary learning algorithm can learn atoms from noisy graph data. Experimental results demonstrate that our algorithm achieves good performance on both simulated and real-world datasets. | This paper proposes a robust graph dictionary learning method based on a novel robust variant of GWD. |
Temporal data like time series are often observed at irregular intervals which is a challenging setting for the existing machine learning methods. To tackle this problem, we view such data as samples from some underlying continuous function. We then define a diffusion-based generative model that adds noise from a predefined stochastic process while preserving the continuity of the resulting underlying function. A neural network is trained to reverse this process which allows us to sample new realizations from the learned distribution. We define suitable stochastic processes as noise sources and introduce novel denoising and score-matching models on processes. Further, we show how to apply this approach to the multivariate probabilistic forecasting and imputation tasks. Through our extensive experiments, we demonstrate that our method outperforms previous models on synthetic and real-world datasets. | We modify the diffusion framework to model continuous functions and apply the learned generative model on different time series tasks. |
Adversarial training (AT) is the de facto method to build robust neural networks, but it is computationally expensive. To overcome this, fast single-step attacks can be used, but doing so is prone to catastrophic overfitting (CO). This is when networks gain non-trivial robustness during the first stages of AT, but then reach a breaking point where they become vulnerable in just a few iterations. Although some works have succeeded at preventing CO, the different mechanisms that lead to this failure mode are still poorly understood. In this work, we study the onset of CO in single-step AT methods through controlled modifications of typical datasets of natural images. In particular, we show that CO can be induced when injecting the images with seemingly innocuous features that are very useful for non-robust classification but need to be combined with other features to obtain a robust classifier. This new perspective provides important insights into the mechanisms that lead to CO and improves our understanding of the general dynamics of adversarial training. | We analyse the phenomena of catastrophic overfitting trough active interventions and show it is a shortcut for the network to avoid learning complex robust solutions. |
Gradient-based multilevel optimization (MLO) has gained attention as a framework for studying numerous problems, ranging from hyperparameter optimization and meta-learning to neural architecture search and reinforcement learning. However, gradients in MLO, which are obtained by composing best-response Jacobians via the chain rule, are notoriously difficult to implement and memory/compute intensive. We take an initial step towards closing this gap by introducing Betty, a software library for large-scale MLO. At its core, we devise a novel dataflow graph for MLO, which allows us to (1) develop efficient automatic differentiation for MLO that reduces the computational complexity from $\mathcal{O}(d^3)$ to $\mathcal{O}(d^2)$, (2) incorporate systems support such as mixed-precision and data-parallel training for scalability, and (3) facilitate implementation of MLO programs of arbitrary complexity while allowing a modular interface for diverse algorithmic and systems design choices. We empirically demonstrate that Betty can be used to implement an array of MLO programs, while also observing up to 11% increase in test accuracy, 14% decrease in GPU memory usage, and 20% decrease in training wall time over existing implementations on multiple benchmarks. We also showcase that Betty enables scaling MLO to models with hundreds of millions of parameters. We open-source the code at https://github.com/leopard-ai/betty. | We develop a scalable, user-friendly, and modular automatic differentiation library for multilevel optimization based on a novel interpretation of multilevel optimization as a dataflow graph. |
A good model for action-effect prediction, i.e., the environment model, is essential for sample-efficient policy learning, in which the agent can take numerous free trials to find good policies. Currently, the model is commonly learned by fitting historical transition data through empirical risk minimization (ERM). However, we discover that simple data fitting can lead to a model that will be totally wrong in guiding policy learning due to the selection bias in offline dataset collection. In this work, we introduce weighted empirical risk minimization (WERM) to handle this problem in model learning. A typical WERM method utilizes inverse propensity scores to re-weight the training data to approximate the target distribution. However, during the policy training, the data distributions of the candidate policies can be various and unknown. Thus, we propose an adversarial weighted empirical risk minimization (AWRM) objective that learns the model with respect to the worst case of the target distributions. We implement AWRM in a sequential decision structure, resulting in the GALILEO model learning algorithm. We also discover that GALILEO is closely related to adversarial model learning, explaining the empirical effectiveness of the latter. We apply GALILEO in synthetic tasks and verify that GALILEO makes accurate predictions on counterfactual data. We finally applied GALILEO in real-world offline policy learning tasks and found that GALILEO significantly improves policy performance in real-world testing. | We propose a new environment model learning techniques with better genalization ability on counterfactual data. |