Datasets:

Rahmaa

/

SciTLDR_ClEaN

like 0

Logit Regularization Methods for Adversarial Robustness

While great progress has been made at making neural networks effective across a wide range of tasks, many are surprisingly vulnerable to small, carefully chosen perturbations of their input, known as adversarial examples.In this paper, we advocate for and experimentally investigate the use of logit regularization techniques as an adversarial defense, which can be used in conjunction with other methods for creating adversarial robustness at little to no cost.We demonstrate that much of the effectiveness of one recent adversarial defense mechanism can be attributed to logit regularization and show how to improve its defense against both white-box and black-box attacks, in the process creating a stronger black-box attacks against PGD-based models.

Logit regularization methods help explain and improve state of the art adversarial defenses

DeepArchitect: Automatically Designing and Training Deep Architectures

In deep learning, performance is strongly affected by the choice of architectureand hyperparameters.While there has been extensive work on automatic hyperpa-rameter optimization for simple spaces, complex spaces such as the space of deeparchitectures remain largely unexplored.As a result, the choice of architecture isdone manually by the human expert through a slow trial and error process guidedmainly by intuition.In this paper we describe a framework for automaticallydesigning and training deep models.We propose an extensible and modular lan-guage that allows the human expert to compactly represent complex search spacesover architectures and their hyperparameters.The resulting search spaces are tree-structured and therefore easy to traverse.Models can be automatically compiled tocomputational graphs once values for all hyperparameters have been chosen.Wecan leverage the structure of the search space to introduce different model searchalgorithms, such as random search, Monte Carlo tree search, and sequen-tial model-based optimization.We present experiments comparing thedifferent algorithms on CIFAR-10 and show that MCTS and SMBO outperformrandom search.We also present experiments on MNIST, showing that the samesearch space achieves near state-of-the-art performance with a few samples.Theseexperiments show that our framework can be used effectively for model discov-ery, as it is possible to describe expressive search spaces and discover competitivemodels without much effort from the human expert.Code for our framework andexperiments has been made publicly available

We describe a modular and composable language for describing expressive search spaces over architectures and simple model search algorithms applied to these search spaces.

Apprentice: Using Knowledge Distillation Techniques To Improve Low-Precision Network Accuracy

Deep learning networks have achieved state-of-the-art accuracies on computer vision workloads like image classification and object detection.The performant systems, however, typically involve big models with numerous parameters.Once trained, a challenging aspect for such top performing models is deployment on resource constrained inference systems -- the models are compute and memory intensive.Low precision numerics and model compression using knowledge distillation are popular techniques to lower both the compute requirements and memory footprint of these deployed models.In this paper, we study the combination of these two techniques and show that the performance of low precision networks can be significantly improved by using knowledge distillation techniques.We call our approach Apprentice and show state-of-the-art accuracies using ternary precision and 4-bit precision for many variants of ResNet architecture on ImageNet dataset.We study three schemes in which one can apply knowledge distillation techniques to various stages of the train-and-deploy pipeline.

We show that knowledge transfer techniques can improve the accuracy of low precision networks and set new state-of-the-art accuracy for ternary and 4-bits precision.

BNN+: Improved Binary Network Training

Deep neural networks are widely used in many applications.However, their deployment on edge devices has been difficult because they are resource hungry.Binary neural networks help to alleviate the prohibitive resource requirements of DNN, where both activations and weights are limited to 1-bit.We propose an improved binary training method, by introducing a regularization function that encourages training weights around binary values.In addition to this, to enhance model performance we add trainable scaling factors to our regularization functions.Furthermore, we use an improved approximation of the derivative of the sign activation function in the backward computation.These additions are based on linear operations that are easily implementable into the binary training framework.We show experimental results on CIFAR-10 obtaining an accuracy of 86.5%, on AlexNet and 91.3% with VGG network.On ImageNet, our method also outperforms the traditional BNN method and XNOR-net, using AlexNet by a margin of 4% and 2% top-1 accuracy respectively.

The paper presents an improved training mechanism for obtaining binary networks with smaller accuracy drop that helps close the gap with it's full precision counterpart

HYPE: Human-eYe Perceptual Evaluation of Generative Models

Generative models often use human evaluations to determine and justify progress.Unfortunately, existing human evaluation methods are ad-hoc: there is currently no standardized, validated evaluation that: measures perceptual fidelity, is reliable, separates models into clear rank order, and ensures high-quality measurement without intractable cost.In response, we construct Human-eYe Perceptual Evaluation, a human metric that is grounded in psychophysics research in perception, reliable across different sets of randomly sampled outputs from a model, results in separable model performances, and efficient in cost and time.We introduce two methods.The first, HYPE-Time, measures visual perception under adaptive time constraints to determine the minimum length of time that model output such as a generated face needs to be visible for people to distinguish it as real or fake.The second, HYPE-Infinity, measures human error rate on fake and real images with no time constraints, maintaining stability and drastically reducing time and cost.We test HYPE across four state-of-the-art generative adversarial networks on unconditional image generation using two datasets, the popular CelebA and the newer higher-resolution FFHQ, and two sampling techniques of model outputs."By simulating HYPE's evaluation multiple times, we demonstrate consistent ranking of different models, identifying StyleGAN with truncation trick sampling as superior to StyleGAN without truncation on FFHQ.

HYPE is a reliable human evaluation metric for scoring generative models, starting with human face generation across 4 GANs.

Unsupervised Machine Translation Using Monolingual Corpora Only

Machine translation has recently achieved impressive performance thanks to recent advances in deep learning and the availability of large-scale parallel corpora.There have been numerous attempts to extend these successes to low-resource language pairs, yet requiring tens of thousands of parallel sentences.In this work, we take this research direction to the extreme and investigate whether it is possible to learn to translate even without any parallel data.We propose a model that takes sentences from monolingual corpora in two different languages and maps them into the same latent space.By learning to reconstruct in both languages from this shared feature space, the model effectively learns to translate without using any labeled data.We demonstrate our model on two widely used datasets and two language pairs, reporting BLEU scores of 32.8 and 15.1 on the Multi30k and WMT English-French datasets, without using even a single parallel sentence at training time.

We propose a new unsupervised machine translation model that can learn without using parallel corpora; experimental results show impressive performance on multiple corpora and pairs of languages.

Intrinsic Social Motivation via Causal Influence in Multi-Agent RL

We derive a new intrinsic social motivation for multi-agent reinforcement learning, in which agents are rewarded for having causal influence over another agent's actions, where causal influence is assessed using counterfactual reasoning.", "The reward does not depend on observing another agent's reward function, and is thus a more realistic approach to MARL than taken in previous work.", "We show that the causal influence reward is related to maximizing the mutual information between agents' actions.", 'We test the approach in challenging social dilemma environments, where it consistently leads to enhanced cooperation between agents and higher collective reward.Moreover, we find that rewarding influence can lead agents to develop emergent communication protocols.Therefore, we also employ influence to train agents to use an explicit communication channel, and find that it leads to more effective communication and higher collective reward.Finally, we show that influence can be computed by equipping each agent with an internal model that predicts the actions of other agents.This allows the social influence reward to be computed without the use of a centralised controller, and as such represents a significantly more general and scalable inductive bias for MARL with independent agents.

We reward agents for having a causal influence on the actions of other agents, and show that this gives rise to better cooperation and more meaningful emergent communication protocols.

Distributed Distributional Deterministic Policy Gradients

This work adopts the very successful distributional perspective on reinforcement learning and adapts it to the continuous control setting.We combine this within a distributed framework for off-policy learning in order to develop what we call the Distributed Distributional Deep Deterministic Policy Gradient algorithm, D4PG.We also combine this technique with a number of additional, simple improvements such as the use of N-step returns and prioritized experience replay.Experimentally we examine the contribution of each of these individual components, and show how they interact, as well as their combined contributions.Our results show that across a wide variety of simple control tasks, difficult manipulation tasks, and a set of hard obstacle-based locomotion tasks the D4PG algorithm achieves state of the art performance.

We develop an agent that we call the Distributional Deterministic Deep Policy Gradient algorithm, which achieves state of the art performance on a number of challenging continuous control problems.

Learning Gaussian Policies from Smoothed Action Value Functions

State-action value functions are ubiquitous in reinforcement learning, giving rise to popular algorithms such as SARSA and Q-learning.We propose a new notion of action value defined by a Gaussian smoothed version of the expected Q-value used in SARSA.We show that such smoothed Q-values still satisfy a Bellman equation, making them naturally learnable from experience sampled from an environment.Moreover, the gradients of expected reward with respect to the mean and covariance of a parameterized Gaussian policy can be recovered from the gradient and Hessian of the smoothed Q-value function.Based on these relationships we develop new algorithms for training a Gaussian policy directly from a learned Q-value approximator.The approach is also amenable to proximal optimization techniques by augmenting the objective with a penalty on KL-divergence from a previous policy.We find that the ability to learn both a mean and covariance during training allows this approach to achieve strong results on standard continuous control benchmarks.

We propose a new Q-value function that enables better learning of Gaussian policies.

Graph Constrained Reinforcement Learning for Natural Language Action Spaces

Interactive Fiction games are text-based simulations in which an agent interacts with the world purely through natural language.They are ideal environments for studying how to extend reinforcement learning agents to meet the challenges of natural language understanding, partial observability, and action generation in combinatorially-large text-based action spaces.We present KG-A2C, an agent that builds a dynamic knowledge graph while exploring and generates actions using a template-based action space.We contend that the dual uses of the knowledge graph to reason about game state and to constrain natural language generation are the keys to scalable exploration of combinatorially large natural language actions.Results across a wide variety of IF games show that KG-A2C outperforms current IF agents despite the exponential increase in action space size.

We present KG-A2C, a reinforcement learning agent that builds a dynamic knowledge graph while exploring and generates natural language using a template-based action space - outperforming all current agents on a wide set of text-based games.

The power of deeper networks for expressing natural functions

It is well-known that neural networks are universal approximators, but that deeper networks tend in practice to be more powerful than shallower ones.We shed light on this by proving that the total number of neurons m required to approximate natural classes of multivariate polynomials of n variables grows only linearly with n for deep neural networks, but grows exponentially when merely a single hidden layer is allowed.We also provide evidence that when the number of hidden layers is increased from 1 to k, the neuron requirement grows exponentially not with n but with n^, suggesting that the minimum number of layers required for practical expressibility grows only logarithmically with n.

We prove that deep neural networks are exponentially more efficient than shallow ones at approximating sparse multivariate polynomials.

Neuron ranking - an informed way to compress convolutional neural networks

Convolutional neural networks in recent years have made a dramatic impact in science, technology and industry, yet the theoretical mechanism of CNN architecture design remains surprisingly vague.The CNN neurons, including its distinctive element, convolutional filters, are known to be learnable features, yet their individual role in producing the output is rather unclear.The thesis of this work is that not all neurons are equally important and some of them contain more useful information to perform a given task.Hence, we propose to quantify and rank neuron importance, and directly incorporate neuron importance in the objective function under two formulations: a game theoretical approach based on Shapley value which computes the marginal contribution of each filter; and a probabilistic approach based on what-we-call, the importance switch using variational inference.Using these two methods we confirm the general theory that some of the neurons are inherently more important than the others.Various experiments illustrate that learned ranks can be readily useable for structured network compression and interpretability of learned features.

We propose CNN neuron ranking with two different methods and show their consistency in producing the result which allows to interpret what network deems important and compress the network by keeping the most relevant nodes.

Learning To Explore Using Active Neural Mapping

This work presents a modular and hierarchical approach to learn policies for exploring 3D environments.Our approach leverages the strengths of both classical and learning-based methods, by using analytical path planners with learned mappers, and global and local policies.Use of learning provides flexibility with respect to input modalities, leverages structural regularities of the world, and provides robustness to errors in state estimation.Such use of learning within each module retains its benefits, while at the same time, hierarchical decomposition and modular training allow us to sidestep the high sample complexities associated with training end-to-end policies.Our experiments in visually and physically realistic simulated 3D environments demonstrate the effectiveness of our proposed approach over past learning and geometry-based approaches.

A modular and hierarchical approach to learn policies for exploring 3D environments.

SPIGAN: Privileged Adversarial Learning from Simulation

Deep Learning for Computer Vision depends mainly on the source of supervision.Photo-realistic simulators can generate large-scale automatically labeled synthetic data, but introduce a domain gap negatively impacting performance.We propose a new unsupervised domain adaptation algorithm, called SPIGAN, relying on Simulator Privileged Information and Generative Adversarial Networks.We use internal data from the simulator as PI during the training of a target task network.We experimentally evaluate our approach on semantic segmentation.We train the networks on real-world Cityscapes and Vistas datasets, using only unlabeled real-world images and synthetic labeled data with z-buffer PI from the SYNTHIA dataset.Our method improves over no adaptation and state-of-the-art unsupervised domain adaptation techniques.

An unsupervised sim-to-real domain adaptation method for semantic segmentation using privileged information from a simulator with GAN-based image translation.

Intriguing Properties of Adversarial Training at Scale

Adversarial training is one of the main defenses against adversarial attacks.In this paper, we provide the first rigorous study on diagnosing elements of large-scale adversarial training on ImageNet, which reveals two intriguing properties.First, we study the role of normalization.Batch normalization is a crucial element for achieving state-of-the-art performance on many vision tasks, but we show it may prevent networks from obtaining strong robustness in adversarial training.One unexpected observation is that, for models trained with BN, simply removing clean images from training data largely boosts adversarial robustness, i.e., 18.3%.We relate this phenomenon to the hypothesis that clean images and adversarial images are drawn from two different domains.This two-domain hypothesis may explain the issue of BN when training with a mixture of clean and adversarial images, as estimating normalization statistics of this mixture distribution is challenging.Guided by this two-domain hypothesis, we show disentangling the mixture distribution for normalization, i.e., applying separate BNs to clean and adversarial images for statistics estimation, achieves much stronger robustness.Additionally, we find that enforcing BNs to behave consistently at training and testing can further enhance robustness.Second, we study the role of network capacity.We find our so-called "deep" networks are still shallow for the task of adversarial learning.Unlike traditional classification tasks where accuracy is only marginally improved by adding more layers to "deep" networks, adversarial training exhibits a much stronger demand on deeper networks to achieve higher adversarial robustness.This robustness improvement can be observed substantially and consistently even by pushing the network capacity to an unprecedented scale, i.e., ResNet-638.

The first rigor diagnose of large-scale adversarial training on ImageNet

Bias Also Matters: Bias Attribution for Deep Neural Network Explanation

The gradient of a deep neural network w.r.t. the input provides information that can be used to explain the output prediction in terms of the input features and has been widely studied to assist in interpreting DNNs. In a linear model=wx+bwbbwg=wx+bw$ and the bias attribution, providing separate and complementary explanations.We study several possible attribution methods applied to the bias of each layer in BBp.In experiments, we show that BBp can generate complementary and highly interpretable explanations of DNNs in addition to gradient-based attributions.

Attribute the bias terms of deep neural networks to input features by a backpropagation-type algorithm; Generate complementary and highly interpretable explanations of DNNs in addition to gradient-based attributions.

A fully automated periodicity detection in time series

This paper presents a method to autonomously find periodicities in a signal.It is based on the same idea of using Fourier Transform and autocorrelation function presented in Vlachos et al. 2005.While showing interesting results this method does not perform well on noisy signals or signals with multiple periodicities.Thus, our method adds several new extra steps to fix these issues.Experimental results show that the proposed method outperforms the state of the art algorithms.

This paper presents a method to autonomously find multiple periodicities in a signal, using FFT and ACF and add three news steps (clustering/filtering/detrending)

Adversarial Exploration Strategy for Self-Supervised Imitation Learning

We present an adversarial exploration strategy, a simple yet effective imitation learning scheme that incentivizes exploration of an environment without any extrinsic reward or human demonstration.Our framework consists of a deep reinforcement learning agent and an inverse dynamics model contesting with each other.The former collects training samples for the latter, and its objective is to maximize the error of the latter.The latter is trained with samples collected by the former, and generates rewards for the former when it fails to predict the actual action taken by the former.In such a competitive setting, the DRL agent learns to generate samples that the inverse dynamics model fails to predict correctly, and the inverse dynamics model learns to adapt to the challenging samples.We further propose a reward structure that ensures the DRL agent collects only moderately hard samples and not overly hard ones that prevent the inverse model from imitating effectively.We evaluate the effectiveness of our method on several OpenAI gym robotic arm and hand manipulation tasks against a number of baseline models.Experimental results show that our method is comparable to that directly trained with expert demonstrations, and superior to the other baselines even without any human priors.

A simple yet effective imitation learning scheme that incentivizes exploration of an environment without any extrinsic reward or human demonstration.

DUAL SPACE LEARNING WITH VARIATIONAL AUTOENCODERS

This paper proposes a dual variational autoencoder, a framework for generating images corresponding to multiclass labels.Recent research on conditional generative models, such as the Conditional VAE, exhibit image transfer by changing labels.However, when the dimension of multiclass labels is large, these models cannot change images corresponding to labels, because learning multiple distributions of the corresponding class is necessary to transfer an image.This leads to the lack of training data.Therefore, instead of conditioning with labels, we condition with latent vectors that include label information.DualVAE divides one distribution of the latent space by linear decision boundaries using labels.Consequently, DualVAE can easily transfer an image by moving a latent vector toward a decision boundary and is robust to the missing values of multiclass labels.To evaluate our proposed method, we introduce a conditional inception score for measuring how much an image changes to the target class.We evaluate the images transferred by DualVAE using the CIS in CelebA datasets and demonstrate state-of-the-art performance in a multiclass setting.

a new framework using dual space for generating images corresponding to multiclass labels when the number of class is large

Graph Classification with Geometric Scattering

One of the most notable contributions of deep learning is the application of convolutional neural networks to structured signal classification, and in particular image classification.Beyond their impressive performances in supervised learning, the structure of such networks inspired the development of deep filter banks referred to as scattering transforms.These transforms apply a cascade of wavelet transforms and complex modulus operators to extract features that are invariant to group operations and stable to deformations.Furthermore, ConvNets inspired recent advances in geometric deep learning, which aim to generalize these networks to graph data by applying notions from graph signal processing to learn deep graph filter cascades.We further advance these lines of research by proposing a geometric scattering transform using graph wavelets defined in terms of random walks on the graph.We demonstrate the utility of features extracted with this designed deep filter bank in graph classification of biochemistry and social network data, and in data exploration, where they enable inference of EC exchange preferences in enzyme evolution.

We present a new feed forward graph ConvNet based on generalizing the wavelet scattering transform of Mallat, and demonstrate its utility in graph classification and data exploration tasks.

CORD: A Consolidated Receipt Dataset for Post-OCR Parsing

OCR is inevitably linked to NLP since its final output is in text.Advances in document intelligence are driving the need for a unified technology that integrates OCR with various NLP tasks, especially semantic parsing.Since OCR and semantic parsing have been studied as separate tasks so far, the datasets for each task on their own are rich, while those for the integrated post-OCR parsing tasks are relatively insufficient.In this study, we publish a consolidated dataset for receipt parsing as the first step towards post-OCR parsing tasks.The dataset consists of thousands of Indonesian receipts, which contains images and box/text annotations for OCR, and multi-level semantic labels for parsing.The proposed dataset can be used to address various OCR and parsing tasks.

We introduce a large-scale receipt dataset for post-OCR parsing tasks.

Pipelined Training with Stale Weights of Deep Convolutional Neural Networks

The growth in the complexity of Convolutional Neural Networks is increasing interest in partitioning a network across multiple accelerators during training and pipelining the backpropagation computations over the accelerators.Existing approaches avoid or limit the use of stale weights through techniques such as micro-batching or weight stashing.These techniques either underutilize of accelerators or increase memory footprint.We explore the impact of stale weights on the statistical efficiency and performance in a pipelined backpropagation scheme that maximizes accelerator utilization and keeps memory overhead modest.We use 4 CNNs and show that when pipelining is limited to early layers in a network, training with stale weights converges and results in models with comparable inference accuracies to those resulting from non-pipelined training on MNIST and CIFAR-10 datasets; a drop in accuracy of 0.4%, 4%, 0.83% and 1.45% for the 4 networks, respectively.However, when pipelining is deeper in the network, inference accuracies drop significantly.We propose combining pipelined and non-pipelined training in a hybrid scheme to address this drop.We demonstrate the implementation and performance of our pipelined backpropagation in PyTorch on 2 GPUs using ResNet, achieving speedups of up to 1.8X over a 1-GPU baseline, with a small drop in inference accuracy.

Accelerating CNN training on a Pipeline of Accelerators with Stale Weights

Excessive Invariance Causes Adversarial Vulnerability

Despite their impressive performance, deep neural networks exhibit striking failures on out-of-distribution inputs.One core idea of adversarial example research is to reveal neural network errors under such distribution shifts.We decompose these errors into two complementary sources: sensitivity and invariance.We show deep networks are not only too sensitive to task-irrelevant changes of their input, as is well-known from epsilon-adversarial examples, but are also too invariant to a wide range of task-relevant changes, thus making vast regions in input space vulnerable to adversarial attacks.We show such excessive invariance occurs across various tasks and architecture types.On MNIST and ImageNet one can manipulate the class-specific content of almost any image without changing the hidden activations.We identify an insufficiency of the standard cross-entropy loss as a reason for these failures.Further, we extend this objective based on an information-theoretic analysis so it encourages the model to consider all task-dependent features in its decision.This provides the first approach tailored explicitly to overcome excessive invariance and resulting vulnerabilities.

We show deep networks are not only too sensitive to task-irrelevant changes of their input, but also too invariant to a wide range of task-relevant changes, thus making vast regions in input space vulnerable to adversarial attacks.

Flow++: Improving Flow-Based Generative Models with Variational Dequantization and Architecture Design

Flow-based generative models are powerful exact likelihood models with efficient sampling and inference.Despite their computational efficiency, flow-based models generally have much worse density modeling performance compared to state-of-the-art autoregressive models.In this paper, we investigate and improve upon three limiting design choices employed by flow-based models in prior work: the use of uniform noise for dequantization, the use of inexpressive affine flows, and the use of purely convolutional conditioning networks in coupling layers.Based on our findings, we propose Flow++, a new flow-based model that is now the state-of-the-art non-autoregressive model for unconditional density estimation on standard image benchmarks.Our work has begun to close the significant performance gap that has so far existed between autoregressive models and flow-based models.

Improved training of current flow-based generative models (Glow and RealNVP) on density estimation benchmarks

Data augmentation instead of explicit regularization

Modern deep artificial neural networks have achieved impressive results through models with orders of magnitude more parameters than training examples which control overfitting with the help of regularization.Regularization can be implicit, as is the case of stochastic gradient descent and parameter sharing in convolutional layers, or explicit.Explicit regularization techniques, most common forms are weight decay and dropout, have proven successful in terms of improved generalization, but they blindly reduce the effective capacity of the model, introduce sensitive hyper-parameters and require deeper and wider architectures to compensate for the reduced capacity.In contrast, data augmentation techniques exploit domain knowledge to increase the number of training examples and improve generalization without reducing the effective capacity and without introducing model-dependent parameters, since it is applied on the training data.In this paper we systematically contrast data augmentation and explicit regularization on three popular architectures and three data sets.Our results demonstrate that data augmentation alone can achieve the same performance or higher as regularized models and exhibits much higher adaptability to changes in the architecture and the amount of training data.

Deep neural networks trained with data augmentation do not require any other explicit regularization (such as weight decay and dropout) and exhibit greater adaptaibility to changes in the architecture and the amount of training data.

Censoring Representations with Multiple-Adversaries over Random Subspaces

Adversarial feature learning is one of the promising ways for explicitly constrains neural networks to learn desired representations; for example, AFL could help to learn anonymized representations so as to avoid privacy issues.AFL learn such a representations by training the networks to deceive the adversary that predict the sensitive information from the network, and therefore, the success of the AFL heavily relies on the choice of the adversary.This paper proposes a novel design of the adversary, that instantiate the concept of the.The proposed method is motivated by an assumption that deceiving an adversary could fail to give meaningful information if the adversary is easily fooled, and adversary rely on single classifier suffer from this issues.In contrast, the proposed method is designed to be less vulnerable, by utilizing the ensemble of independent classifiers where each classifier tries to predict sensitive variables from a different of the representations.The empirical validations on three user-anonymization tasks show that our proposed method achieves state-of-the-art performances in all three datasets without significantly harming the utility of data.This is significant because it gives new implications about designing the adversary, which is important to improve the performance of AFL.

This paper improves the quality of the recently proposed adversarial feature leaning (AFL) approach for incorporating explicit constrains to representations, by introducing the concept of the of the adversary.

Variational inference of latent hierarchical dynamical systems in neuroscience: an application to calcium imaging data

A key problem in neuroscience, and life sciences more generally, is that data is generated by a hierarchy of dynamical systems.One example of this is in calcium imaging data, where data is generated by a lower-order dynamical system governing calcium flux in neurons, which itself is driven by a higher-order dynamical system of neural computation.Ideally, life scientists would be able to infer the dynamics of both the lower-order systems and the higher-order systems, but this is difficult in high-dimensional regimes.A recent approach using sequential variational auto-encoders demonstrated it was possible to learn the latent dynamics of a single dynamical system for computations during reaching behaviour in the brain, using spiking data modelled as a Poisson process.Here we extend this approach using a ladder method to infer a hierarchy of dynamical systems, allowing us to capture calcium dynamics as well as neural computation.In this approach, spiking events drive lower-order calcium dynamics, and are themselves controlled by a higher-order latent dynamical system.We generate synthetic data by generating firing rates, sampling spike trains, and converting spike trains to fluorescence transients, from two dynamical systems that have been used as key benchmarks in recent literature: a Lorenz attractor, and a chaotic recurrent neural network.We show that our model is better able to reconstruct Lorenz dynamics from fluorescence data than competing methods.However, though our model can reconstruct underlying spike rates and calcium transients from the chaotic neural network well, it does not perform as well at reconstructing firing rates as basic techniques for inferring spikes from calcium data.These results demonstrate that VLAEs are a promising approach for modelling hierarchical dynamical systems data in the life sciences, but that inferring the dynamics of lower-order systems can potentially be better achieved with simpler methods.

We extend a successful recurrent variational autoencoder for dynamic systems to model an instance of dynamic systems hierarchy in neuroscience using the ladder method.

Linear Symmetric Quantization of Neural Networks for Low-precision Integer Hardware

With the proliferation of specialized neural network processors that operate on low-precision integers, the performance of Deep Neural Network inference becomes increasingly dependent on the result of quantization.Despite plenty of prior work on the quantization of weights or activations for neural networks, there is still a wide gap between the software quantizers and the low-precision accelerator implementation, which degrades either the efficiency of networks or that of the hardware for the lack of software and hardware coordination at design-phase.In this paper, we propose a learned linear symmetric quantizer for integer neural network processors, which not only quantizes neural parameters and activations to low-bit integer but also accelerates hardware inference by using batch normalization fusion and low-precision accumulators and multipliers.We use a unified way to quantize weights and activations, and the results outperform many previous approaches for various networks such as AlexNet, ResNet, and lightweight models like MobileNet while keeping friendly to the accelerator architecture.Additional, we also apply the method to object detection models and witness high performance and accuracy in YOLO-v2.Finally, we deploy the quantized models on our specialized integer-arithmetic-only DNN accelerator to show the effectiveness of the proposed quantizer.We show that even with linear symmetric quantization, the results can be better than asymmetric or non-linear methods in 4-bit networks.In evaluation, the proposed quantizer induces less than 0.4% accuracy drop in ResNet18, ResNet34, and AlexNet when quantizing the whole network as required by the integer processors.

We introduce an efficient quantization process that allows for performance acceleration on specialized integer-only neural network accelerator.

EnergyNet: Energy-Efficient Dynamic Inference

The prohibitive energy cost of running high-performance Convolutional Neural Networks has been limiting their deployment on resource-constrained platforms including mobile and wearable devices.We propose a CNN for energy-aware dynamic routing, called the EnergyNet, that achieves adaptive-complexity inference based on the inputs, leading to an overall reduction of run time energy cost without noticeably losing accuracy.That is achieved by proposing an energy loss that captures both computational and data movement costs.We combine it with the accuracy-oriented loss, and learn a dynamic routing policy for skipping certain layers in the networks, that optimizes the hybrid loss. Our empirical results demonstrate that, compared to the baseline CNNs, EnergyNetcan trim down the energy cost up to 40% and 65%, during inference on the CIFAR10 and Tiny ImageNet testing sets, respectively, while maintaining the same testing accuracies. It is further encouraging to observe that the energy awareness might serve as a training regularization and can even improve prediction accuracy: our models can achieve 0.7% higher top-1 testing accuracy than the baseline on CIFAR-10 when saving up to 27% energy, and 1.0% higher top-5 testing accuracy on Tiny ImageNet when saving up to 50% energy, respectively.

This paper proposes a new CNN model that combines energy cost with a dynamic routing strategy to enable adaptive energy-efficient inference.

LSH Softmax: Sub-Linear Learning and Inference of the Softmax Layer in Deep Architectures

Log-linear models models are widely used in machine learning, and in particular are ubiquitous in deep learning architectures in the form of the softmax.While exact inference and learning of these requires linear time, it can be done approximately in sub-linear time with strong concentrations guarantees.In this work, we present LSH Softmax, a method to perform sub-linear learning and inference of the softmax layer in the deep learning setting.Our method relies on the popular Locality-Sensitive Hashing to build a well-concentrated gradient estimator, using nearest neighbors and uniform samples.We also present an inference scheme in sub-linear time for LSH Softmax using the Gumbel distribution.On language modeling, we show that Recurrent Neural Networks trained with LSH Softmax perform on-par with computing the exact softmax while requiring sub-linear computations.

we present LSH Softmax, a softmax approximation layer for sub-linear learning and inference with strong theoretical guarantees; we showcase both its applicability and efficiency by evaluating on a real-world task: language modeling.

Learning scalable and transferable multi-robot/machine sequential assignment planning via graph embedding

Can the success of reinforcement learning methods for simple combinatorial optimization problems be extended to multi-robot sequential assignment planning?In addition to the challenge of achieving near-optimal performance in large problems, transferability to an unseen number of robots and tasks is another key challenge for real-world applications.In this paper, we suggest a method that achieves the first success in both challenges for robot/machine scheduling problems. Our method comprises of three components.First, we show any robot scheduling problem can be expressed as a random probabilistic graphical model.We develop a mean-field inference method for random PGM and use it for Q-function inference.Second, we show that transferability can be achieved by carefully designing two-step sequential encoding of problem state.Third, we resolve the computational scalability issue of fitted Q-iteration by suggesting a heuristic auction-based Q-iteration fitting method enabled by transferability we achieved. We apply our method to discrete-time, discrete space problems) and scalably achieve 97% optimality with transferability.This optimality is maintained under stochastic contexts.By extending our method to continuous time, continuous space formulation, we claim to be the first learning-based method with scalable performance in any type of multi-machine scheduling problems; our method scalability achieves comparable performance to popular metaheuristics in Identical parallel machine scheduling problems.

RL can solve (stochastic) multi-robot/scheduling problems scalably and transferably using graph embedding

Learning of Sophisticated Curriculums by viewing them as Graphs over Tasks

Curriculum learning consists in learning a difficult task by first training on an easy version of it, then on more and more difficult versions and finally on the difficult task.To make this learning efficient, given a curriculum and the current learning state of an agent, we need to find what are the good next tasks to train the agent on.Teacher-Student algorithms assume that the good next tasks are the ones on which the agent is making the fastest progress or digress.We first simplify and improve them."However, two problematic situations where the agent is mainly trained on tasks it can't learn yet or it already learnt may occur.", 'Therefore, we introduce a new algorithm using min max ordered curriculums that assumes that the good next tasks are the ones that are learnable but not learnt yet.It outperforms Teacher-Student algorithms on small curriculums and significantly outperforms them on sophisticated ones with numerous tasks.

We present a new algorithm for learning by curriculum based on the notion of mastering rate that outperforms previous algorithms.

Neocortical plasticity: an unsupervised cake but no free lunch

The fields of artificial intelligence and neuroscience have a long history of fertile bi-directional interactions.On the one hand, important inspiration for the development of artificial intelligence systems has come from the study of natural systems of intelligence, the mammalian neocortex in particular.On the other, important inspiration for models and theories of the brain have emerged from artificial intelligence research.A central question at the intersection of these two areas is concerned with the processes by which neocortex learns, and the extent to which they are analogous to the back-propagation training algorithm of deep networks.Matching the data efficiency, transfer and generalisation properties of neocortical learning remains an area of active research in the field of deep learning.Recent advances in our understanding of neuronal, synaptic and dendritic physiology of the neocortex suggest new approaches for unsupervised representation learning, perhaps through a new class of objective functions, which could act alongside or in lieu of back-propagation.Such local learning rules have implicit rather than explicit objectives with respect to the training data, facilitating domain adaptation and generalisation. Incorporating them into deep networks for representation learning could better leverage unlabelled datasets to offer significant improvements in data efficiency of downstream supervised readout learning, and reduce susceptibility to adversarial perturbations, at the cost of a more restricted domain of applicability.

Inspiration from local dendritic processes of neocortical learning to make unsupervised learning great again.

Towards Language Agnostic Universal Representations

When a bilingual student learns to solve word problems in math, we expect the student to be able to solve these problem in both languages the student is fluent in, even if the math lessons were only taught in one language.However, current representations in machine learning are language dependent.In this work, we present a method to decouple the language from the problem by learning language agnostic representations and therefore allowing training a model in one language and applying to a different one in a zero shot fashion.We learn these representations by taking inspiration from linguistics, specifically the Universal Grammar hypothesis and learn universal latent representations that are language agnostic.We demonstrate the capabilities of these representations by showing that the models trained on a single language using language agnostic representations achieve very similar accuracies in other languages.

By taking inspiration from linguistics, specifically the Universal Grammar hypothesis, we learn language agnostic universal representations which we can utilize to do zero-shot learning across languages.

Improving Gaussian mixture latent variable model convergence with Optimal Transport

Generative models with both discrete and continuous latent variables are highly motivated by the structure of many real-world data sets.They present, however, subtleties in training often manifesting in the discrete latent variable not being leveraged.In this paper, we show why such models struggle to train using traditional log-likelihood maximization, and that they are amenable to training using the Optimal Transport framework of Wasserstein Autoencoders.We find our discrete latent variable to be fully leveraged by the model when trained, without any modifications to the objective function or significant fine tuning.Our model generates comparable samples to other approaches while using relatively simple neural networks, since the discrete latent variable carries much of the descriptive burden.Furthermore, the discrete latent provides significant control over generation.

This paper shows that the Wasserstein distance objective enables the training of latent variable models with discrete latents in a case where the Variational Autoencoder objective fails to do so.

An Attention-Based Model for Learning Dynamic Interaction Networks

While machine learning models achieve human-comparable performance on sequential data, exploiting structured knowledge is still a challenging problem.Spatio-temporal graphs have been proved to be a useful tool to abstract interaction graphs and previous works exploits carefully designed feed-forward architecture to preserve such structure.We argue to scale such network design to real-world problem, a model needs to automatically learn a meaningful representation of the possible relations.Learning such interaction structure is not trivial: on the one hand, a model has to discover the hidden relations between different problem factors in an unsupervised way; on the other hand, the mined relations have to be interpretable.In this paper, we propose an attention module able to project a graph sub-structure in a fixed size embedding, preserving the influence that the neighbours exert on a given vertex.On a comprehensive evaluation done on real-world as well as toy task, we found our model competitive against strong baselines.

A graph neural network able to automatically learn and leverage a dynamic interactive graph structure

NAMSG: An Efficient Method for Training Neural Networks

We introduce NAMSG, an adaptive first-order algorithm for training neural networks.The method is efficient in computation and memory, and is straightforward to implement.It computes the gradients at configurable remote observation points, in order to expedite the convergence by adjusting the step size for directions with different curvatures in the stochastic setting.It also scales the updating vector elementwise by a nonincreasing preconditioner to take the advantages of AMSGRAD.We analyze the convergence properties for both convex and nonconvex problems by modeling the training process as a dynamic system, and provide a strategy to select the observation factor without grid search.A data-dependent regret bound is proposed to guarantee the convergence in the convex setting.The method can further achieve a O) regret bound for strongly convex functions.Experiments demonstrate that NAMSG works well in practical problems and compares favorably to popular adaptive methods, such as ADAM, NADAM, and AMSGRAD.

A new algorithm for training neural networks that compares favorably to popular adaptive methods.

Dissecting an Adversarial framework for Information Retrieval

Recent advances in Generative Adversarial Networks facilitated by improvements to the framework and successful application to various problems has resulted in extensions to multiple domains.IRGAN attempts to leverage the framework for Information-Retrieval, a task that can be described as modeling the correct conditional probability distribution p over the documents, given the query.The work that proposes IRGAN claims that optimizing their minimax loss function will result in a generator which can learn the distribution, but their setup and baseline term steer the model away from an exact adversarial formulation, and this work attempts to point out certain inaccuracies in their formulation.Analyzing their loss curves gives insight into possible mistakes in the loss functions and better performance can be obtained by using the co-training like setup we propose, where two models are trained in a co-operative rather than an adversarial fashion.

Points out problems in loss function used in IRGAN, a recently proposed GAN framework for Information Retrieval. Further, a model motivated by co-training is proposed, which achieves better performance.

Federated User Representation Learning

Collaborative personalization, such as through learned user representations, can improve the prediction accuracy of neural-network-based models significantly.We propose Federated User Representation Learning, a simple, scalable, privacy-preserving and resource-efficient way to utilize existing neural personalization techniques in the Federated Learning setting.FURL divides model parameters into federated and private parameters.Private parameters, such as private user embeddings, are trained locally, but unlike federated parameters, they are not transferred to or averaged on the server.We show theoretically that this parameter split does not affect training for most model personalization approaches.Storing user embeddings locally not only preserves user privacy, but also improves memory locality of personalization compared to on-server training.We evaluate FURL on two datasets, demonstrating a significant improvement in model quality with 8% and 51% performance increases, and approximately the same level of performance as centralized training with only 0% and 4% reductions.Furthermore, we show that user embeddings learned in FL and the centralized setting have a very similar structure, indicating that FURL can learn collaboratively through the shared parameters while preserving user privacy.

We propose Federated User Representation Learning (FURL), a simple, scalable, privacy-preserving and bandwidth-efficient way to utilize existing neural personalization techniques in the Federated Learning (FL) setting.

Discrete Autoencoders for Sequence Models

Recurrent models for sequences have been recently successful at many tasks, especially for language modelingand machine translation.Nevertheless, it remains challenging to extract good representations fromthese models.For instance, even though language has a clear hierarchical structure going from charactersthrough words to sentences, it is not apparent in current language models.We propose to improve the representation in sequence models byaugmenting current approaches with an autoencoder that is forced to compressthe sequence through an intermediate discrete latent space.In order to propagate gradientsthough this discrete representation we introduce an improved semantic hashing technique.We show that this technique performs well on a newly proposed quantitative efficiency measure.We also analyze latent codes produced by the model showing how they correspond towords and phrases.Finally, we present an application of the autoencoder-augmentedmodel to generating diverse translations.

Autoencoders for text with a new method for using discrete latent space.

Unaligned Image-to-Sequence Transformation with Loop Consistency

We tackle the problem of modeling sequential visual phenomena.Given examples of a phenomena that can be divided into discrete time steps, we aim to take an input from any such time and realize this input at all other time steps in the sequence.Furthermore, we aim to do this ground-truth aligned sequences --- avoiding the difficulties needed for gathering aligned data.This generalizes the unpaired image-to-image problem from generating pairs to generating sequences.We extend cycle consistency to and alleviate difficulties associated with learning in the resulting long chains of computation."We show competitive results compared to existing image-to-image techniques when modeling several different data sets including the Earth's seasons and aging of human faces.

LoopGAN extends cycle length in CycleGAN to enable unaligned sequential transformation for more than two time steps.

Stochastic Weight Averaging in Parallel: Large-Batch Training That Generalizes Well

We propose Stochastic Weight Averaging in Parallel, an algorithm to accelerate DNN training.Our algorithm uses large mini-batches to compute an approximate solution quickly and then refines it by averaging the weights of multiple models computed independently and in parallel.The resulting models generalize equally well as those trained with small mini-batches but are produced in a substantially shorter time.We demonstrate the reduction in training time and the good generalization performance of the resulting models on the computer vision datasets CIFAR10, CIFAR100, and ImageNet.

We propose SWAP, a distributed algorithm for large-batch training of neural networks.

Large Batch Training of Convolutional Networks with Layer-wise Adaptive Rate Scaling

A common way to speed up training of large convolutional networks is to add computational units.Training is then performed using data-parallel synchronous Stochastic Gradient Descent with a mini-batch divided between computational units.With an increase in the number of nodes, the batch size grows.However, training with a large batch often results in lower model accuracy.We argue that the current recipe for large batch training is not general enough and training may diverge.To overcome these optimization difficulties, we propose a new training algorithm based on Layer-wise Adaptive Rate Scaling.Using LARS, we scaled AlexNet and ResNet-50 to a batch size of 16K.

A new large batch training algorithm based on Layer-wise Adaptive Rate Scaling (LARS); using LARS, we scaled AlexNet and ResNet-50 to a batch of 16K.

Learning Compositional Koopman Operators for Model-Based Control

Finding an embedding space for a linear approximation of a nonlinear dynamical system enables efficient system identification and control synthesis.The Koopman operator theory lays the foundation for identifying the nonlinear-to-linear coordinate transformations with data-driven methods.Recently, researchers have proposed to use deep neural networks as a more expressive class of basis functions for calculating the Koopman operators.These approaches, however, assume a fixed dimensional state space; they are therefore not applicable to scenarios with a variable number of objects.In this paper, we propose to learn compositional Koopman operators, using graph neural networks to encode the state into object-centric embeddings and using a block-wise linear transition matrix to regularize the shared structure across objects.The learned dynamics can quickly adapt to new environments of unknown physical parameters and produce control signals to achieve a specified goal.Our experiments on manipulating ropes and controlling soft robots show that the proposed method has better efficiency and generalization ability than existing baselines.

Learning compositional Koopman operators for efficient system identification and model-based control.

Scalable Gradients and Variational Inference for Stochastic Differential Equations

We derive reverse-mode automatic differentiation for solutions of stochastic differential equations, allowing time-efficient and constant-memory computation of pathwise gradients, a continuous-time analogue of the reparameterization trick.Specifically, we construct a backward SDE whose solution is the gradient and provide conditions under which numerical solutions converge.We also combine our stochastic adjoint approach with a stochastic variational inference scheme for continuous-time SDE models, allowing us to learn distributions over functions using stochastic gradient descent.Our latent SDE model achieves competitive performance compared to existing approaches on time series modeling.

We present a constant memory gradient computation procedure through solutions of stochastic differential equations (SDEs) and apply the method for learning latent SDE models.

Coordinated Exploration via Intrinsic Rewards for Multi-Agent Reinforcement Learning

Solving tasks with sparse rewards is one of the most important challenges in reinforcement learning.In the single-agent setting, this challenge has been addressed by introducing intrinsic rewards that motivate agents to explore unseen regions of their state spaces.Applying these techniques naively to the multi-agent setting results in agents exploring independently, without any coordination among themselves.We argue that learning in cooperative multi-agent settings can be accelerated and improved if agents coordinate with respect to what they have explored.In this paper we propose an approach for learning how to dynamically select between different types of intrinsic rewards which consider not just what an individual agent has explored, but all agents, such that the agents can coordinate their exploration and maximize extrinsic returns.Concretely, we formulate the approach as a hierarchical policy where a high-level controller selects among sets of policies trained on different types of intrinsic rewards and the low-level controllers learn the action policies of all agents under these specific rewards.We demonstrate the effectiveness of the proposed approach in a multi-agent gridworld domain with sparse rewards, and then show that our method scales up to more complex settings by evaluating on the VizDoom platform.

We propose several intrinsic reward functions for encouraging coordinated exploration in multi-agent problems, and introduce an approach to dynamically selecting the best exploration method for a given task, online.

Learning Classifier Synthesis for Generalized Few-Shot Learning

Object recognition in real-world requires handling long-tailed or even open-ended data.An ideal visual system needs to reliably recognize the populated visual concepts and meanwhile efficiently learn about emerging new categories with a few training instances.Class-balanced many-shot learning and few-shot learning tackle one side of this problem, via either learning strong classifiers for populated categories or learning to learn few-shot classifiers for the tail classes.In this paper, we investigate the problem of generalized few-shot learning -- a model during the deployment is required to not only learn about "tail" categories with few shots, but simultaneously classify the "head" and "tail" categories."We propose the Classifier Synthesis Learning, a learning framework that learns how to synthesize calibrated few-shot classifiers in addition to the multi-class classifiers of head classes, leveraging a shared neural dictionary.", 'CASTLE sheds light upon the inductive GFSL through optimizing one clean and effective GFSL learning objective.It demonstrates superior performances than existing GFSL algorithms and strong baselines on MiniImageNet and TieredImageNet data sets.More interestingly, it outperforms previous state-of-the-art methods when evaluated on standard few-shot learning.

We propose to learn synthesizing few-shot classifiers and many-shot classifiers using one single objective function for GFSL.

Knossos: Compiling AI with AI

Machine learning workloads are often expensive to train, taking weeks to converge.The current generation of frameworks relies on custom back-ends in order to achieve efficiency, making it impractical to train models on less common hardware where no such back-ends exist.Knossos builds on recent work that avoids the need for hand-written libraries, instead compiles machine learning models in much the same way one would compile other kinds of software.In order to make the resulting code efficient, the Knossos complier directly optimises the abstract syntax tree of the program.However in contrast to traditional compilers that employ hand-written optimisation passes, we take a rewriting approach driven by the search algorithm and a learn value function that evaluates future potential cost reduction of taking various rewriting actions to the program.We show that Knossos can automatically learned optimisations that past compliers had to implement by hand.Furthermore, we demonstrate that Knossos can achieve wall time reduction compared to a hand-tuned compiler on a suite of machine learning programs, including basic linear algebra and convolutional networks.The Knossos compiler has minimal dependencies and can be used on any architecture that supports a \\Cpp toolchain.Since cost model the proposed algorithm optimises can be tailored to a particular hardware architecture, the proposed approach can potentially applied to a variety of hardware.

We combine A* search with reinforcement learning to speed up machine learning code

Data-efficient Deep Reinforcement Learning for Dexterous Manipulation

Grasping an object and precisely stacking it on another is a difficult task for traditional robotic control or hand-engineered approaches.Here we examine the problem in simulation and provide techniques aimed at solving it via deep reinforcement learning.We introduce two straightforward extensions to the Deep Deterministic Policy Gradient algorithm, which make it significantly more data-efficient and scalable.Our results show that by making extensive use of off-policy data and replay, it is possible to find high-performance control policies.Further, our results hint that it may soon be feasible to train successful stacking policies by collecting interactions on real robots.

Data-efficient deep reinforcement learning can be used to learning precise stacking policies.

Policy Gradient For Multidimensional Action Spaces: Action Sampling and Entropy Bonus

In recent years deep reinforcement learning has been shown to be adept at solving sequential decision processes with high-dimensional state spaces such as in the Atari games.Many reinforcement learning problems, however, involve high-dimensional discrete action spaces as well as high-dimensional state spaces.In this paper, we develop a novel policy gradient methodology for the case of large multidimensional discrete action spaces.We propose two approaches for creating parameterized policies: LSTM parameterization and a Modified MDP giving rise to Feed-Forward Network parameterization.Both of these approaches provide expressive models to which backpropagation can be applied for training.We then consider entropy bonus, which is typically added to the reward function to enhance exploration.In the case of high-dimensional action spaces, calculating the entropy and the gradient of the entropy requires enumerating all the actions in the action space and running forward and backpropagation for each action, which may be computationally infeasible.We develop several novel unbiased estimators for the entropy bonus and its gradient.Finally, we test our algorithms on two environments: a multi-hunter multi-rabbit grid game and a multi-agent multi-arm bandit problem.

policy parameterizations and unbiased policy entropy estimators for MDP with large multidimensional discrete action space

Asynchronous Stochastic Subgradient Methods for General Nonsmooth Nonconvex Optimization

Asynchronous distributed methods are a popular way to reduce the communication and synchronization costs of large-scale optimization.Yet, for all their success, little is known about their convergence guarantees in the challenging case of general non-smooth, non-convex objectives, beyond cases where closed-form proximal operator solutions are available.This is all the more surprising since these objectives are the ones appearing in the training of deep neural networks.In this paper, we introduce the first convergence analysis covering asynchronous methods in the case of general non-smooth, non-convex objectives.Our analysis applies to stochastic sub-gradient descent methods both with and without block variable partitioning, and both with and without momentum.It is phrased in the context of a general probabilistic model of asynchronous scheduling accurately adapted to modern hardware properties.We validate our analysis experimentally in the context of training deep neural network architectures.We show their overall successful asymptotic convergence as well as exploring how momentum, synchronization, and partitioning all affect performance.

Asymptotic convergence for stochastic subgradien method with momentum under general parallel asynchronous computation for general nonconvex nonsmooth optimization

Low Bias Gradient Estimates for Very Deep Boolean Stochastic Networks

Stochastic neural networks with discrete random variables are an important class of models for their expressivity and interpretability.Since direct differentiation and backpropagation is not possible, Monte Carlo gradient estimation techniques have been widely employed for training such models.Efficient stochastic gradient estimators, such Straight-Through and Gumbel-Softmax, work well for shallow models with one or two stochastic layers.Their performance, however, suffers with increasing model complexity.In this work we focus on stochastic networks with multiple layers of Boolean latent variables.To analyze such such networks, we employ the framework of harmonic analysis for Boolean functions. We use it to derive an analytic formulation for the source of bias in the biased Straight-Through estimator.Based on the analysis we propose , a simple gradient estimation algorithm that relies on three simple bias reduction steps.Extensive experiments show that FouST performs favorably compared to state-of-the-art biased estimators, while being much faster than unbiased ones.To the best of our knowledge FouST is the first gradient estimator to train up very deep stochastic neural networks, with up to 80 deterministic and 11 stochastic layers.

We present a low-bias estimator for Boolean stochastic variable models with many stochastic layers.

ManiGAN: Text-Guided Image Manipulation

We propose a novel generative adversarial network for visual attributes manipulation, which is able to semantically modify the visual attributes of given images using natural language descriptions.The key to our method is to design a novel co-attention module to combine text and image information rather than simply concatenating two features along the channel direction.Also, a detail correction module is proposed to rectify mismatched attributes of the synthetic image, and to reconstruct text-unrelated contents.Finally, we propose a new metric for evaluating manipulation results, in terms of both the generation of text-related attributes and the reconstruction of text-unrelated contents.Extensive experiments on benchmark datasets demonstrate the advantages of our proposed method, regarding the effectiveness of image manipulation and the capability of generating high-quality results.

We propose a novel method to manipulate given images using natural language descriptions.

Probabilistic Planning with Sequential Monte Carlo methods

In this work, we propose a novel formulation of planning which views it as a probabilistic inference problem over future optimal trajectories.This enables us to use sampling methods, and thus, tackle planning in continuous domains using a fixed computational budget. We design a new algorithm, Sequential Monte Carlo Planning, by leveraging classical methods in Sequential Monte Carlo and Bayesian smoothing in the context of control as inference.Furthermore, we show that Sequential Monte Carlo Planning can capture multimodal policies and can quickly learn continuous control tasks.

Leveraging control as inference and Sequential Monte Carlo methods, we proposed a probabilistic planning algorithm.

On Self Modulation for Generative Adversarial Networks

Training Generative Adversarial Networks is notoriously challenging.We propose and study an architectural modification, self-modulation, which improves GAN performance across different data sets, architectures, losses, regularizers, and hyperparameter settings.Intuitively, self-modulation allows the intermediate feature maps of a generator to change as a function of the input noise vector.While reminiscent of other conditioning techniques, it requires no labeled data.In a large-scale empirical study we observe a relative decrease of 5%-35% in FID.Furthermore, all else being equal, adding this modification to the generator leads to improved performance in 124/144 of the studied settings.Self-modulation is a simple architectural change that requires no additional parameter tuning, which suggests that it can be applied readily to any GAN.

A simple GAN modification that improves performance across many losses, architectures, regularization schemes, and datasets.

A Deep Dive into Count-Min Sketch for Extreme Classification in Logarithmic Memory

Extreme Classification Methods have become of paramount importance, particularly for Information Retrieval problems, owing to the development of smart algorithms that are scalable to industry challenges.One of the prime class of models that aim to solve the memory and speed challenge of extreme multi-label learning is Group Testing.Multi-label Group Testing methods construct label groups by grouping original labels either randomly or based on some similarity and then train smaller classifiers to first predict the groups and then recover the original label vectors.Recently, a novel approach called MACH was proposed which projects the huge label vectors to a small and manageable count-min sketch matrix and then learns to predict this matrix to recover the original prediction probabilities.Thereby, the model memory scales O for K classes.MACH is a simple algorithm which works exceptionally well in practice.Despite this simplicity of MACH, there is a big gap between the theoretical understanding of the trade-offs with MACH.In this paper we fill this gap.Leveraging the theory of count-min sketch we provide precise quantification of the memory-identifiablity tradeoffs.We extend the theory to the case of multi-label classification, where the dependencies make the estimators hard to calculate in closed forms.To mitigate this issue, we propose novel quadratic approximation using the Inclusion-Exclusion Principle.Our estimator has significantly lower reconstruction error than the typical CMS estimator across various values of number of classes K, label sparsity and compression ratio.

How to estimate original probability vector for millions of classes from count-min sketch measurements - a theoretical and practical setup.

Fix your classifier: the marginal value of training the last weight layer

Neural networks are commonly used as models for classification for a wide variety of tasks.Typically, a learned affine transformation is placed at the end of such models, yielding a per-class value used for classification.This classifier can have a vast number of parameters, which grows linearly with the number of possible classes, thus requiring increasingly more resources.In this work we argue that this classifier can be fixed, up to a global scale constant, with little or no loss of accuracy for most tasks, allowing memory and computational benefits.Moreover, we show that by initializing the classifier with a Hadamard matrix we can speed up inference as well.We discuss the implications for current understanding of neural network models.

You can fix the classifier in neural networks without losing accuracy

Found by NEMO: Unsupervised Object Detection from Negative Examples and Motion

This paper introduces NEMO, an approach to unsupervised object detection that uses motion---instead of image labels---as a cue to learn object detection.To discriminate between motion of the target object and other changes in the image, it relies on negative examples that show the scene without the object.The required data can be collected very easily by recording two short videos, a positive one showing the object in motion and a negative one showing the scene without the object.Without any additional form of pretraining or supervision and despite of occlusions, distractions, camera motion, and adverse lighting, those videos are sufficient to learn object detectors that can be applied to new videos and even generalize to unseen scenes and camera angles.In a baseline comparison, unsupervised object detection outperforms off-the shelf template matching and tracking approaches that are given an initial bounding box of the object.The learned object representations are also shown to be accurate enough to capture the relevant information from manipulation task demonstrations, which makes them applicable to learning from demonstration in robotics.An example of object detection that was learned from 3 minutes of video can be found here: http://y2u.be/u_jyz9_ETz4

Learning to detect objects without image labels from 3 minutes of video

ReClor: A Reading Comprehension Dataset Requiring Logical Reasoning

Recent powerful pre-trained language models have achieved remarkable performance on most of the popular datasets for reading comprehension.It is time to introduce more challenging datasets to push the development of this field towards more comprehensive reasoning of text.In this paper, we introduce a new Reading Comprehension dataset requiring logical reasoning extracted from standardized graduate admission examinations.As earlier studies suggest, human-annotated datasets usually contain biases, which are often exploited by models to achieve high accuracy without truly understanding the text.In order to comprehensively evaluate the logical reasoning ability of models on ReClor, we propose to identify biased data points and separate them into EASY set while the rest as HARD set.Empirical results show that the state-of-the-art models have an outstanding ability to capture biases contained in the dataset with high accuracy on EASY set.However, they struggle on HARD set with poor performance near that of random guess, indicating more research is needed to essentially enhance the logical reasoning ability of current models.

We introduce ReClor, a reading comprehension dataset requiring logical reasoning, and find that current state-of-the-art models struggle with real logical reasoning with poor performance near that of random guess.

Garbage in, model out: Weight theft with just noise

This paper explores the scenarios under whichan attacker can claim that ‘Noise and access tothe softmax layer of the model is all you need’to steal the weights of a convolutional neural networkwhose architecture is already known.Wewere able to achieve 96% test accuracy usingthe stolen MNIST model and 82% accuracy usingstolen KMNIST model learned using onlyi.i.d. Bernoulli noise inputs.We posit that thistheft-susceptibility of the weights is indicativeof the complexity of the dataset and propose anew metric that captures the same.The goal ofthis dissemination is to not just showcase how farknowing the architecture can take you in terms ofmodel stealing, but to also draw attention to thisrather idiosyncratic weight learnability aspects ofCNNs spurred by i.i.d. noise input.We also disseminatesome initial results obtained with usingthe Ising probability distribution in lieu of the i.i.d.Bernoulli distribution

Input only noise , glean the softmax outputs, steal the weights

Implicit λ-Jeffreys Autoencoders: Taking the Best of Both Worlds

We propose a new form of an autoencoding model which incorporates the best properties of variational autoencoders and generative adversarial networks.It is known that GAN can produce very realistic samples while VAE does not suffer from mode collapsing problem.Our model optimizes λ-Jeffreys divergence between the model distribution and the true data distribution.We show that it takes the best properties of VAE and GAN objectives.It consists of two parts.One of these parts can be optimized by using the standard adversarial training, and the second one is the very objective of the VAE model.However, the straightforward way of substituting the VAE loss does not work well if we use an explicit likelihood such as Gaussian or Laplace which have limited flexibility in high dimensions and are unnatural for modelling images in the space of pixels.To tackle this problem we propose a novel approach to train the VAE model with an implicit likelihood by an adversarially trained discriminator.In an extensive set of experiments on CIFAR-10 and TinyImagent datasets, we show that our model achieves the state-of-the-art generation and reconstruction quality and demonstrate how we can balance between mode-seeking and mode-covering behaviour of our model by adjusting the weight λ in our objective.

We propose a new form of an autoencoding model which incorporates the best properties of variational autoencoders (VAE) and generative adversarial networks (GAN)

Modulating transfer between tasks in gradient-based meta-learning

Learning-to-learn or meta-learning leverages data-driven inductive bias to increase the efficiency of learning on a novel task.This approach encounters difficulty when transfer is not mutually beneficial, for instance, when tasks are sufficiently dissimilar or change over time.Here, we use the connection between gradient-based meta-learning and hierarchical Bayes to propose a mixture of hierarchical Bayesian models over the parameters of an arbitrary function approximator such as a neural network.Generalizing the model-agnostic meta-learning algorithm, we present a stochastic expectation maximization procedure to jointly estimate parameter initializations for gradient descent as well as a latent assignment of tasks to initializations.This approach better captures the diversity of training tasks as opposed to consolidating inductive biases into a single set of hyperparameters.Our experiments demonstrate better generalization on the standard miniImageNet benchmark for 1-shot classification.We further derive a novel and scalable non-parametric variant of our method that captures the evolution of a task distribution over time as demonstrated on a set of few-shot regression tasks.

We use the connection between gradient-based meta-learning and hierarchical Bayes to learn a mixture of meta-learners that is appropriate for a heterogeneous and evolving task distribution.

Capsules with Inverted Dot-Product Attention Routing

We introduce a new routing algorithm for capsule networks, in which a child capsule is routed to a parent based only on agreement between the parent's state and the child's vote.", "Unlike previously proposed routing algorithms, the parent's ability to reconstruct the child is not explicitly taken into account to update the routing probabilities.", 'This simplifies the routing procedure and improves performance on benchmark datasets such as CIFAR-10 and CIFAR-100.The new mechanism 1) designs routing via inverted dot-product attention; 2) imposes Layer Normalization as normalization; and 3) replaces sequential iterative routing with concurrent iterative routing.Besides outperforming existing capsule networks, our model performs at-par with a powerful CNN, using less than 25% of the parameters. On a different task of recognizing digits from overlayed digit images, the proposed capsule model performs favorably against CNNs given the same number of layers and neurons per layer. We believe that our work raises the possibility of applying capsule networks to complex real-world tasks.

We present a new routing method for Capsule networks, and it performs at-par with ResNet-18 on CIFAR-10/ CIFAR-100.

Doc2Dial: a Framework for Dialogue Composition Grounded in Business Documents

We introduce Doc2Dial, an end-to-end framework for generating conversational data grounded in business documents via crowdsourcing.Such data can be used to train automated dialogue agents performing customer care tasks for the enterprises or organizations.In particular, the framework takes the documents as input and generates the tasks for obtaining the annotations for simulating dialog flows.The dialog flows are used to guide the collection of utterances produced by crowd workers.The outcomes include dialogue data grounded in the given documents, as well as various types of annotations that help ensure the quality of the data and the flexibility tocomposite dialogues.

We introduce Doc2Dial, an end-to-end framework for generating conversational data grounded in business documents via crowdsourcing for train automated dialogue agents

MelNet: A Generative Model for Audio in the Frequency Domain

Capturing high-level structure in audio waveforms is challenging because a single second of audio spans tens of thousands of timesteps. While long-range dependencies are difficult to model directly in the time domain, we show that they can be more tractably modelled in two-dimensional time-frequency representations such as spectrograms. By leveraging this representational advantage, in conjunction with a highly expressive probabilistic model and a multiscale generation procedure, we design a model capable of generating high-fidelity audio samples which capture structure at timescales which time-domain models have yet to achieve. We demonstrate that our model captures longer-range dependencies than time-domain models such as WaveNet across a diverse set of unconditional generation tasks, including single-speaker speech generation, multi-speaker speech generation, and music generation.

We introduce an autoregressive generative model for spectrograms and demonstrate applications to speech and music generation

Why do deep convolutional networks generalize so poorly to small image transformations?

Deep convolutional network architectures are often assumed to guarantee generalization for small image translations and deformations.In this paper we show that modern CNNs can drastically change their output when an image is translated in the image plane by a few pixels, and that this failure of generalization also happens with other realistic small image transformations.Furthermore, we see these failures to generalize more frequently in more modern networks.We show that these failures are related to the fact that the architecture of modern CNNs ignores the classical sampling theorem so that generalization is not guaranteed.We also show that biases in the statistics of commonly used image datasets makes it unlikely that CNNs will learn to be invariant to these transformations.Taken together our results suggest that the performance of CNNs in object recognition falls far short of the generalization capabilities of humans.

Modern deep CNNs are not invariant to translations, scalings and other realistic image transformations, and this lack of invariance is related to the subsampling operation and the biases contained in image datasets.

Auto-Conditioned Recurrent Networks for Extended Complex Human Motion Synthesis

We present a real-time method for synthesizing highly complex human motions using a novel training regime we call the auto-conditioned Recurrent Neural Network.Recently, researchers have attempted to synthesize new motion by using autoregressive techniques, but existing methods tend to freeze or diverge after a couple of seconds due to an accumulation of errors that are fed back into the network.Furthermore, such methods have only been shown to be reliable for relatively simple human motions, such as walking or running.In contrast, our approach can synthesize arbitrary motions with highly complex styles, including dances or martial arts in addition to locomotion.The acRNN is able to accomplish this by explicitly accommodating for autoregressive noise accumulation during training.Our work is the first to our knowledge that demonstrates the ability to generate over 18,000 continuous frames of new complex human motion w.r.t. different styles.

Synthesize complex and extended human motions using an auto-conditioned LSTM network

A Simple Technique to Enable Saliency Methods to Pass the Sanity Checks

attempt to explain a deep net's decision by assigning a to each feature/pixel in the input, often doing this credit-assignment via the gradient of the output with respect to input.", 'Recently t questioned the validity of many of these methods since they do not pass simple, which test whether the scores shift/vanish when layers of the trained net are randomized, or when the net is retrained using random labels for inputs.% for the inputs. %Surprisingly, the tested methods did not pass these checks: the explanations were relatively unchanged.We propose a simple fix to existing saliency methods that helps them pass sanity checks, which we call.This involves computing saliency maps for all possible labels in the classification task, and using a simple competition among them to identify and remove less relevant pixels from the map.Some theoretical justification is provided for it and its performance is empirically demonstrated on several popular methods.

We devise a mechanism called competition among pixels that allows (approximately) complete saliency methods to pass the sanity checks.

Are You Sure YouWant To Do That? Classification with Interpretable Queries

Classification systems typically act in isolation, meaning they are required to implicitly memorize the characteristics of all candidate classes in order to classify.The cost of this is increased memory usage and poor sample efficiency.We propose a model which instead verifies using reference images during the classification process, reducing the burden of memorization.The model uses iterative non-differentiable queries in order to classify an image.We demonstrate that such a model is feasible to train and can match baseline accuracy while being more parameter efficient.However, we show that finding the correct balance between image recognition and verification is essential to pushing the model towards desired behavior, suggesting that a pipeline of recognition followed by verification is a more promising approach towards designing more powerful networks with simpler architectures.

Image classification via iteratively querying for reference image from a candidate class with a RNN and use CNN to compare to the input image

A Main/Subsidiary Network Framework for Simplifying Binary Neural Networks

To reduce memory footprint and run-time latency, techniques such as neural net-work pruning and binarization have been explored separately. However, it is un-clear how to combine the best of the two worlds to get extremely small and efficient models. In this paper, we, for the first time, define the filter-level pruning problem for binary neural networks, which cannot be solved by simply migrating existing structural pruning methods for full-precision models. A novel learning-based approach is proposed to prune filters in our main/subsidiary network frame-work, where the main network is responsible for learning representative features to optimize the prediction performance, and the subsidiary component works as a filter selector on the main network. To avoid gradient mismatch when training the subsidiary component, we propose a layer-wise and bottom-up scheme. We also provide the theoretical and experimental comparison between our learning-based and greedy rule-based methods. Finally, we empirically demonstrate the effectiveness of our approach applied on several binary models, including binarizedNIN, VGG-11, and ResNet-18, on various image classification datasets. For bi-nary ResNet-18 on ImageNet, we use 78.6% filters but can achieve slightly better test error 49.87% than the original model

we define the filter-level pruning problem for binary neural networks for the first time and propose method to solve it.

AntMan: Sparse Low-Rank Compression To Accelerate RNN Inference

Wide adoption of complex RNN based models is hindered by their inference performance, cost and memory requirements.To address this issue, we develop AntMan, combining structured sparsity with low-rank decomposition synergistically, to reduce model computation, size and execution time of RNNs while attaining desired accuracy.AntMan extends knowledge distillation based training to learn the compressed models efficiently.Our evaluation shows that AntMan offers up to 100x computation reduction with less than 1pt accuracy drop for language and machine reading comprehension models.Our evaluation also shows that for a given accuracy target, AntMan produces 5x smaller models than the state-of-art.Lastly, we show that AntMan offers super-linear speed gains compared to theoretical speedup, demonstrating its practical value on commodity hardware.

Reducing computation and memory complexity of RNN models by up to 100x using sparse low-rank compression modules, trained via knowledge distillation.

Residual Gated Graph ConvNets

Graph-structured data such as social networks, functional brain networks, gene regulatory networks, communications networks have brought the interest in generalizing deep learning techniques to graph domains.In this paper, we are interested to design neural networks for graphs with variable length in order to solve learning problems such as vertex classification, graph classification, graph regression, and graph generative tasks.Most existing works have focused on recurrent neural networks to learn meaningful representations of graphs, and more recently new convolutional neural networks have been introduced.In this work, we want to compare rigorously these two fundamental families of architectures to solve graph learning tasks.We review existing graph RNN and ConvNet architectures, and propose natural extension of LSTM and ConvNet to graphs with arbitrary size.Then, we design a set of analytically controlled experiments on two basic graph problems, i.e. subgraph matching and graph clustering, to test the different architectures. Numerical results show that the proposed graph ConvNets are 3-17% more accurate and 1.5-4x faster than graph RNNs.Graph ConvNets are also 36% more accurate than variational techniques.Finally, the most effective graph ConvNet architecture uses gated edges and residuality.Residuality plays an essential role to learn multi-layer architectures as they provide a 10% gain of performance.

We compare graph RNNs and graph ConvNets, and we consider the most generic class of graph ConvNets with residuality.

Complex- and Real-Valued Neural Network Architectures

Complex-value neural networks are not a new concept, however, the use of real-values has often been favoured over complex-values due to difficulties in training and accuracy of results.Existing literature ignores the number of parameters used.We compared complex- and real-valued neural networks using five activation functions.We found that when real and complex neural networks are compared using simple classification tasks, complex neural networks perform equal to or slightly worse than real-value neural networks.However, when specialised architecture is used, complex-valued neural networks outperform real-valued neural networks.Therefore, complex–valued neural networks should be used when the input data is also complex or it can be meaningfully to the complex plane, or when the network architecture uses the structure defined by using complex numbers.

Comparison of complex- and real-valued multi-layer perceptron with respect to the number of real-valued parameters.

CAYLEYNETS: SPECTRAL GRAPH CNNS WITH COMPLEX RATIONAL FILTERS

The rise of graph-structured data such as social networks, regulatory networks, citation graphs, and functional brain networks, in combination with resounding success of deep learning in various applications, has brought the interest in generalizing deep learning models to non-Euclidean domains.In this paper, we introduce a new spectral domain convolutional architecture for deep learning on graphs.The core ingredient of our model is a new class of parametric rational complex functions allowing to efficiently compute spectral filters on graphs that specialize on frequency bands of interest.Our model generates rich spectral filters that are localized in space, scales linearly with the size of the input data for sparsely-connected graphs, and can handle different constructions of Laplacian operators.Extensive experimental results show the superior performance of our approach on spectral image classification, community detection, vertex classification and matrix completion tasks.

A spectral graph convolutional neural network with spectral zoom properties.

FasterSeg: Searching for Faster Real-time Semantic Segmentation

We present FasterSeg, an automatically designed semantic segmentation network with not only state-of-the-art performance but also faster speed than current methods.Utilizing neural architecture search, FasterSeg is discovered from a novel and broader search space integrating multi-resolution branches, that has been recently found to be vital in manually designed segmentation models.To better calibrate the balance between the goals of high accuracy and low latency, we propose a decoupled and fine-grained latency regularization, that effectively overcomes our observed phenomenons that the searched networks are prone to "collapsing" to low-latency yet poor-accuracy models.Moreover, we seamlessly extend FasterSeg to a new collaborative search framework, simultaneously searching for a teacher and a student network in the same single run.The teacher-student distillation further boosts the student model’s accuracy.Experiments on popular segmentation benchmarks demonstrate the competency of FasterSeg.For example, FasterSeg can run over 30% faster than the closest manually designed competitor on Cityscapes, while maintaining comparable accuracy.

We present a real-time segmentation model automatically discovered by a multi-scale NAS framework, achieving 30% faster than state-of-the-art models.

Making Efficient Use of Demonstrations to Solve Hard Exploration Problems

This paper introduces R2D3, an agent that makes efficient use of demonstrations to solve hard exploration problems in partially observable environments with highly variable initial conditions.We also introduce a suite of eight tasks that combine these three properties, and show that R2D3 can solve several of the tasks where other state of the art methods fail to see even a single successful trajectory after tens of billions of steps of exploration.

We introduce R2D3, an agent that makes efficient use of demonstrations to solve hard exploration problems in partially observable environments with highly variable initial conditions.

An Investigation of Memory in Recurrent Neural Networks

We investigate the learned dynamical landscape of a recurrent neural network solving a simple task requiring the interaction of two memory mechanisms: long- and short-term.Our results show that while long-term memory is implemented by asymptotic attractors, sequential recall is now additionally implemented by oscillatory dynamics in a transverse subspace to the basins of attraction of these stable steady states.Based on our observations, we propose how different types of memory mechanisms can coexist and work together in a single neural network, and discuss possible applications to the fields of artificial intelligence and neuroscience.

We investigate how a recurrent neural network successfully learns a task combining long-term memory and sequential recall.

Count-Based Exploration with the Successor Representation

The problem of exploration in reinforcement learning is well-understood in the tabular case and many sample-efficient algorithms are known.Nevertheless, it is often unclear how the algorithms in the tabular setting can be extended to tasks with large state-spaces where generalization is required.Recent promising developments generally depend on problem-specific density models or handcrafted features.In this paper we introduce a simple approach for exploration that allows us to develop theoretically justified algorithms in the tabular case but that also give us intuitions for new algorithms applicable to settings where function approximation is required.Our approach and its underlying theory is based on the substochastic successor representation, a concept we develop here.While the traditional successor representation is a representation that defines state generalization by the similarity of successor states, the substochastic successor representation is also able to implicitly count the number of times each state has been observed.This extension connects two until now disjoint areas of research.We show in traditional tabular domains that our algorithm empirically performs as well as other sample-efficient algorithms.We then describe a deep reinforcement learning algorithm inspired by these ideas and show that it matches the performance of recent pseudo-count-based methods in hard exploration Atari 2600 games.

We propose the idea of using the norm of the successor representation an exploration bonus in reinforcement learning. In hard exploration Atari games, our the deep RL algorithm matches the performance of recent pseudo-count-based methods.

Likelihood Contribution based Multi-scale Architecture for Generative Flows

Deep generative modeling using flows has gained popularity owing to the tractable exact log-likelihood estimation with efficient training and synthesis process.However, flow models suffer from the challenge of having high dimensional latent space, same in dimension as the input space.An effective solution to the above challenge as proposed by Dinh et al. is a multi-scale architecture, which is based on iterative early factorization of a part of the total dimensions at regular intervals.Prior works on generative flows involving a multi-scale architecture perform the dimension factorization based on a static masking.We propose a novel multi-scale architecture that performs data dependent factorization to decide which dimensions should pass through more flow layers.To facilitate the same, we introduce a heuristic based on the contribution of each dimension to the total log-likelihood which encodes the importance of the dimensions.Our proposed heuristic is readily obtained as part of the flow training process, enabling versatile implementation of our likelihood contribution based multi-scale architecture for generic flow models.We present such an implementation for the original flow introduced in Dinh et al., and demonstrate improvements in log-likelihood score and sampling quality on standard image benchmarks.We also conduct ablation studies to compare proposed method with other options for dimension factorization.

Data-dependent factorization of dimensions in a multi-scale architecture based on contribution to the total log-likelihood

Towards better understanding of gradient-based attribution methods for Deep Neural Networks

Understanding the flow of information in Deep Neural Networks is a challenging problem that has gain increasing attention over the last few years.While several methods have been proposed to explain network predictions, there have been only a few attempts to compare them from a theoretical perspective.What is more, no exhaustive empirical comparison has been performed in the past.In this work we analyze four gradient-based attribution methods and formally prove conditions of equivalence and approximation between them.By reformulating two of these methods, we construct a unified framework which enables a direct comparison, as well as an easier implementation.Finally, we propose a novel evaluation metric, called Sensitivity-n and test the gradient-based attribution methods alongside with a simple perturbation-based attribution method on several datasets in the domains of image and text classification, using various network architectures.

Four existing backpropagation-based attribution methods are fundamentally similar. How to assess it?

Stochastic algorithms under single spiked models

We study SGD and Adam for estimating a rank one signal planted in matrix or tensor noise.The extreme simplicity of the problem setup allows us to isolate the effects of various factors: signal to noise ratio, density of critical points, stochasticity and initialization.We observe a surprising phenomenon: Adam seems to get stuck in local minima as soon as polynomially many critical points appear, while SGD escapes those.However, when the number of critical points degenerates to exponentials, then both algorithms get trapped.Theory tells us that at fixed SNR the problem becomes intractable for large and in our experiments SGD does not escape this.We exhibit the benefits of warm starting in those situations.We conclude that in this class of problems, warm starting cannot be replaced by stochasticity in gradients to find the basin of attraction.

SGD and Adam under single spiked model for tensor PCA

Explaining Neural Networks Semantically and Quantitatively

This paper presents a method to explain the knowledge encoded in a convolutional neural network quantitatively and semantically.How to analyze the specific rationale of each prediction made by the CNN presents one of key issues of understanding neural networks, but it is also of significant practical values in certain applications.In this study, we propose to distill knowledge from the CNN into an explainable additive model, so that we can use the explainable model to provide a quantitative explanation for the CNN prediction.We analyze the typical bias-interpreting problem of the explainable model and develop prior losses to guide the learning of the explainable additive model.Experimental results have demonstrated the effectiveness of our method.

This paper presents a method to explain the knowledge encoded in a convolutional neural network (CNN) quantitatively and semantically.

cGANs with Projection Discriminator

We propose a novel, projection based way to incorporate the conditional information into the discriminator of GANs that respects the role of the conditional information in the underlining probabilistic model.This approach is in contrast with most frameworks of conditional GANs used in application today, which use the conditional information by concatenating the conditional vector to the feature vectors.With this modification, we were able to significantly improve the quality of the class conditional image generation on ILSVRC2012 dataset from the current state-of-the-art result, and we achieved this with a single pair of a discriminator and a generator.We were also able to extend the application to super-resolution and succeeded in producing highly discriminative super-resolution images.This new structure also enabled high quality category transformation based on parametric functional transformation of conditional batch normalization layers in the generator.

We propose a novel, projection based way to incorporate the conditional information into the discriminator of GANs that respects the role of the conditional information in the underlining probabilistic model.

The Frechet Distance of training and test distribution predicts the generalization gap

Learning theory tells us that more data is better when minimizing the generalization error of identically distributed training and test sets.However, when training and test distribution differ, this distribution shift can have a significant effect.With a novel perspective on function transfer learning, we are able to lower bound the change of performance when transferring from training to test set with the Wasserstein distance between the embedded training and test set distribution.We find that there is a trade-off affecting performance between how invariant a function is to changes in training and test distribution and how large this shift in distribution is.Empirically across several data domains, we substantiate this viewpoint by showing that test performance correlates strongly with the distance in data distributions between training and test set.Complementary to the popular belief that more data is always better, our results highlight the utility of also choosing a training data distribution that is close to the test data distribution when the learned function is not invariant to such changes.

The Frechet Distance between train and test distribution correlates with the change in performance for functions that are not invariant to the shift.

Spiroplots: a new discrete-time dynamical system to generate curve patterns

We introduce a new procedural dynamic system that can generate a variety of shapes that often appear as curves, but technically, the figures are plots of many points.We name them spiroplots and show how this new system relates to other procedures or processes that generate figures.Spiroplots are an extremely simple process but with a surprising visual variety.We prove some fundamental properties and analyze some instances to see how the geometry or topology of the input determines the generated figures.We show that some spiroplots have a finite cycle and return to the initial situation, whereas others will produce new points infinitely often.This paper is accompanied by a JavaScript app that allows anyone to generate spiroplots.

A new, very simple dynamic system is introduced that generates pretty patterns; properties are proved and possibilities are explored

GMM-UNIT: Unsupervised Multi-Domain and Multi-Modal Image-to-Image Translation via Attribute Gaussian Mixture Modelling

Unsupervised image-to-image translation aims to learn a mapping between several visual domains by using unpaired training pairs.Recent studies have shown remarkable success in image-to-image translation for multiple domains but they suffer from two main limitations: they are either built from several two-domain mappings that are required to be learned independently and/or they generate low-diversity results, a phenomenon known as model collapse.To overcome these limitations, we propose a method named GMM-UNIT based on a content-attribute disentangled representation, where the attribute space is fitted with a GMM.Each GMM component represents a domain, and this simple assumption has two prominent advantages.First, the dimension of the attribute space does not grow linearly with the number of domains, as it is the case in the literature.Second, the continuous domain encoding allows for interpolation between domains and for extrapolation to unseen domains.Additionally, we show how GMM-UNIT can be constrained down to different methods in the literature, meaning that GMM-UNIT is a unifying framework for unsupervised image-to-image translation.

GMM-UNIT is an image-to-image translation model that maps an image to multiple domains in a stochastic fashion.

Compositional Attention Networks for Machine Reasoning

We present Compositional Attention Networks, a novel fully differentiable neural network architecture, designed to facilitate explicit and expressive reasoning.While many types of neural networks are effective at learning and generalizing from massive quantities of data, this model moves away from monolithic black-box architectures towards a design that provides a strong prior for iterative reasoning, enabling it to support explainable and structured learning, as well as generalization from a modest amount of data.The model builds on the great success of existing recurrent cells such as LSTMs: It sequences a single recurrent Memory, Attention, and Control cell, and by careful design imposes structural constraints on the operation of each cell and the interactions between them, incorporating explicit control and soft attention mechanisms into their interfaces."We demonstrate the model's strength and robustness on the challenging CLEVR dataset for visual reasoning, achieving a new state-of-the-art 98.9% accuracy, halving the error rate of the previous best model.", 'More importantly, we show that the new model is more computationally efficient, data-efficient, and requires an order of magnitude less time and/or data to achieve good results.

We present a novel architecture, based on dynamic memory, attention and composition for the task of machine reasoning.

Imagining the Latent Space of a Variational Auto-Encoders

Variational Auto-Encoders are designed to capture compressible information about a dataset. As a consequence the information stored in the latent space is seldom sufficient to reconstruct a particular image. To help understand the type of information stored in the latent space we train a GAN-style decoder constrained to produce images that the VAE encoder will map to the same region of latent space."This allows us to imagine the information captured in the latent space.We argue that this is necessary to make a VAE into a truly generative model. We use our GAN to visualise the latent space of a standard VAE and of a-VAE.

To understand the information stored in the latent space, we train a GAN-style decoder constrained to produce images that the VAE encoder will map to the same region of latent space.

Hallucinating brains with artificial brains

Human brain function as measured by functional magnetic resonance imaging, exhibits a rich diversity.In response, understanding the individual variabilityof brain function and its association with behavior has become one of themajor concerns in modern cognitive neuroscience.Our work is motivated by theview that generative models provide a useful tool for understanding this variability.To this end, this manuscript presents two novel generative models trainedon real neuroimaging data which synthesize task-dependent functional brain images.Brain images are high dimensional tensors which exhibit structured spatialcorrelations.Thus, both models are 3D conditional Generative Adversarial networks which apply Convolutional Neural Networks to learn anabstraction of brain image representations.Our results show that the generatedbrain images are diverse, yet task dependent.In addition to qualitative evaluation,we utilize the generated synthetic brain volumes as additional training data to improvedownstream fMRI classifiers.Our approach achieves significant improvements for a variety of datasets, classifi-cation tasks and evaluation scores.Our classification results provide a quantitativeevaluation of the quality of the generated images, and also serve as an additionalcontribution of this manuscript.

Two novel GANs are constructed to generate high-quality 3D fMRI brain images and synthetic brain images greatly help to improve downstream classification tasks.

DARTS: Differentiable Architecture Search

This paper addresses the scalability challenge of architecture search by formulating the task in a differentiable manner.Unlike conventional approaches of applying evolution or reinforcement learning over a discrete and non-differentiable search space, our method is based on the continuous relaxation of the architecture representation, allowing efficient search of the architecture using gradient descent.Extensive experiments on CIFAR-10, ImageNet, Penn Treebank and WikiText-2 show that our algorithm excels in discovering high-performance convolutional architectures for image classification and recurrent architectures for language modeling, while being orders of magnitude faster than state-of-the-art non-differentiable techniques.

We propose a differentiable architecture search algorithm for both convolutional and recurrent networks, achieving competitive performance with the state of the art using orders of magnitude less computation resources.

The Curious Case of Neural Text Degeneration

Despite considerable advances in neural language modeling, it remains an open question what the best decoding strategy is for text generation from a language model. The counter-intuitive empirical observation is that even though the use of likelihood as training objective leads to high quality models for a broad range of language understanding tasks, maximization-based decoding methods such as beam search lead to degeneration — output text that is bland, incoherent, or gets stuck in repetitive loops.To address this we propose Nucleus Sampling, a simple but effective method to draw considerably higher quality text out of neural language models.Our approach avoids text degeneration by truncating the unreliable tail of the probability distribution, sampling from the dynamic nucleus of tokens containing the vast majority of the probability mass.To properly examine current maximization-based and stochastic decoding methods, we compare generations from each of these methods to the distribution of human text along several axes such as likelihood, diversity, and repetition.Our results show that maximization is an inappropriate decoding objective for open-ended text generation, the probability distributions of the best current language models have an unreliable tail which needs to be truncated during generation and Nucleus Sampling is the best decoding strategy for generating long-form text that is both high-quality — as measured by human evaluation — and as diverse as human-written text.

Current language generation systems either aim for high likelihood and devolve into generic repetition or miscalibrate their stochasticity—we provide evidence of both and propose a solution: Nucleus Sampling.

Natural Language Adversarial Attack and Defense in Word Level

Up until very recently, inspired by a mass of researches on adversarial examples for computer vision, there has been a growing interest in designing adversarial attacks for Natural Language Processing tasks, followed by very few works of adversarial defenses for NLP.To our knowledge, there exists no defense method against the successful synonym substitution based attacks that aim to satisfy all the lexical, grammatical, semantic constraints and thus are hard to perceived by humans.We contribute to fill this gap and propose a novel adversarial defense method called Synonym Encoding Method, which inserts an encoder before the input layer of the model and then trains the model to eliminate adversarial perturbations.Extensive experiments demonstrate that SEM can efficiently defend current best synonym substitution based adversarial attacks with little decay on the accuracy for benign examples.To better evaluate SEM, we also design a strong attack method called Improved Genetic Algorithm that adopts the genetic metaheuristic for synonym substitution based attacks.Compared with existing genetic based adversarial attack, IGA can achieve higher attack success rate while maintaining the transferability of the adversarial examples.

The first text adversarial defense method in word level, and the improved generic based attack method against synonyms substitution based attacks.

Sparse Attentive Backtracking: Long-Range Credit Assignment in Recurrent Networks

A major drawback of backpropagation through time is the difficulty of learning long-term dependencies, coming from having to propagate credit information backwards through every single step of the forward computation.This makes BPTT both computationally impractical and biologically implausible.For this reason, full backpropagation through time is rarely used on long sequences, and truncated backpropagation through time is used as a heuristic. However, this usually leads to biased estimates of the gradient in which longer term dependencies are ignored. Addressing this issue, we propose an alternative algorithm, Sparse Attentive Backtracking, which might also be related to principles used by brains to learn long-term dependencies.Sparse Attentive Backtracking learns an attention mechanism over the hidden states of the past and selectively backpropagates through paths with high attention weights. This allows the model to learn long term dependencies while only backtracking for a small number of time steps, not just from the recent past but also from attended relevant past states.

Towards Efficient Credit Assignment in Recurrent Networks without Backpropagation Through Time

Learning Discriminators as Energy Networks in Adversarial Learning

We propose a novel adversarial learning framework in this work.Existing adversarial learning methods involve two separate networks, i.e., the structured prediction models and the discriminative models, in the training.The information captured by discriminative models complements that in the structured prediction models, but few existing researches have studied on utilizing such information to improve structured prediction models at the inference stage.In this work, we propose to refine the predictions of structured prediction models by effectively integrating discriminative models into the prediction.Discriminative models are treated as energy-based models.Similar to the adversarial learning, discriminative models are trained to estimate scores which measure the quality of predicted outputs, while structured prediction models are trained to predict contrastive outputs with maximal energy scores.In this way, the gradient vanishing problem is ameliorated, and thus we are able to perform inference by following the ascent gradient directions of discriminative models to refine structured prediction models.The proposed method is able to handle a range of tasks, , multi-label classification and image segmentation. Empirical results on these two tasks validate the effectiveness of our learning method.

We propose a novel adversarial learning framework for structured prediction, in which discriminative models can be used to refine structured prediction models at the inference stage.

RaPP: Novelty Detection with Reconstruction along Projection Pathway

We propose RaPP, a new methodology for novelty detection by utilizing hidden space activation values obtained from a deep autoencoder.Precisely, RaPP compares input and its autoencoder reconstruction not only in the input space but also in the hidden spaces.We show that if we feed a reconstructed input to the same autoencoder again, its activated values in a hidden space are equivalent to the corresponding reconstruction in that hidden space given the original input.In order to aggregate the hidden space activation values, we propose two metrics, which enhance the novelty detection performance.Through extensive experiments using diverse datasets, we validate that RaPP improves novelty detection performances of autoencoder-based approaches.Besides, we show that RaPP outperforms recent novelty detection methods evaluated on popular benchmarks.

A new methodology for novelty detection by utilizing hidden space activation values obtained from a deep autoencoder.

From Language to Goals: Inverse Reinforcement Learning for Vision-Based Instruction Following

Reinforcement learning is a promising framework for solving control problems, but its use in practical situations is hampered by the fact that reward functions are often difficult to engineer.Specifying goals and tasks for autonomous machines, such as robots, is a significant challenge: conventionally, reward functions and goal states have been used to communicate objectives.But people can communicate objectives to each other simply by describing or demonstrating them.How can we build learning algorithms that will allow us to tell machines what we want them to do?In this work, we investigate the problem of grounding language commands as reward functions using inverse reinforcement learning, and argue that language-conditioned rewards are more transferable than language-conditioned policies to new environments.We propose language-conditioned reward learning, which grounds language commands as a reward function represented by a deep neural network.We demonstrate that our model learns rewards that transfer to novel tasks and environments on realistic, high-dimensional visual environments with natural language commands, whereas directly learning a language-conditioned policy leads to poor performance.

We ground language commands in a high-dimensional visual environment by learning language-conditioned rewards using inverse reinforcement learning.

Additive function approximation in the brain

Many biological learning systems such as the mushroom body, hippocampus, and cerebellum are built from sparsely connected networks of neurons.For a new understanding of such networks, we study the function spaces induced by sparse random features and characterize what functions may and may not be learned.A network with d inputs per neuron is found to be equivalent to an additive model of order d, whereas with a degree distribution the network combines additive terms of different orders.We identify three specific advantages of sparsity: additive function approximation is a powerful inductive bias that limits the curse of dimensionality, sparse networks are stable to outlier noise in the inputs, and sparse random features are scalable.Thus, even simple brain architectures can be powerful function approximators.Finally, we hope that this work helps popularize kernel theories of networks among computational neuroscientists.

We advocate for random features as a theory of biological neural networks, focusing on sparsely connected networks

Prob2Vec: Mathematical Semantic Embedding for Problem Retrieval in Adaptive Tutoring

We propose a new application of embedding techniques to problem retrieval in adaptive tutoring.The objective is to retrieve problems similar in mathematical concepts.There are two challenges: First, like sentences, problems helpful to tutoring are never exactly the same in terms of the underlying concepts.Instead, good problems mix concepts in innovative ways, while still displaying continuity in their relationships.Second, it is difficult for humans to determine a similarity score consistent across a large enough training set.We propose a hierarchical problem embedding algorithm, called Prob2Vec, that consists of an abstraction and an embedding step.Prob2Vec achieves 96.88% accuracy on a problem similarity test, in contrast to 75% from directly applying state-of-the-art sentence embedding methods.It is surprising that Prob2Vec is able to distinguish very fine-grained differences among problems, an ability humans need time and effort to acquire.In addition, the sub-problem of concept labeling with imbalanced training data set is interesting in its own right.It is a multi-label problem suffering from dimensionality explosion, which we propose ways to ameliorate.We propose the novel negative pre-training algorithm that dramatically reduces false negative and positive ratios for classification, using an imbalanced training data set.

We propose the Prob2Vec method for problem embedding used in a personalized e-learning tool in addition to a data level classification method, called negative pre-training, for cases where the training data set is imbalanced.

CrescendoNet: A Simple Deep Convolutional Neural Network with Ensemble Behavior

We introduce a new deep convolutional neural network, CrescendoNet, by stacking simple building blocks without residual connections.Each Crescendo block contains independent convolution paths with increased depths.The numbers of convolution layers and parameters are only increased linearly in Crescendo blocks.In experiments, CrescendoNet with only 15 layers outperforms almost all networks without residual connections on benchmark datasets, CIFAR10, CIFAR100, and SVHN.Given sufficient amount of data as in SVHN dataset, CrescendoNet with 15 layers and 4.1M parameters can match the performance of DenseNet-BC with 250 layers and 15.3M parameters.CrescendoNet provides a new way to construct high performance deep convolutional neural networks without residual connections.Moreover, through investigating the behavior and performance of subnetworks in CrescendoNet, we note that the high performance of CrescendoNet may come from its implicit ensemble behavior, which differs from the FractalNet that is also a deep convolutional neural network without residual connections.Furthermore, the independence between paths in CrescendoNet allows us to introduce a new path-wise training procedure, which can reduce the memory needed for training.

We introduce CrescendoNet, a deep CNN architecture by stacking simple building blocks without residual connections.

Deep Random Splines for Point Process Intensity Estimation

Gaussian processes are the leading class of distributions on random functions, but they suffer from well known issues including difficulty scaling and inflexibility with respect to certain shape constraints.Here we propose Deep Random Splines, a flexible class of random functions obtained by transforming Gaussian noise through a deep neural network whose output are the parameters of a spline.Unlike Gaussian processes, Deep Random Splines allow us to readily enforce shape constraints while inheriting the richness and tractability of deep generative models.We also present an observational model for point process data which uses Deep Random Splines to model the intensity function of each point process and apply it to neuroscience data to obtain a low-dimensional representation of spiking activity.Inference is performed via a variational autoencoder that uses a novel recurrent encoder architecture that can handle multiple point processes as input.

We combine splines with neural networks to obtain a novel distribution over functions and use it to model intensity functions of point processes.