Deep Learning Readings We Covered in 2017 Fall Seminar Course

2017Course (Index of Posts):

No. Read Date Title and Information We Read @
1 2017, Aug, 22 Foundations I -Andrew Ng - Nuts and Bolts of Applying Deep Learning 2017-W1
2 2017, Aug, 24 Foundations II - Ganguli - Theoretical Neuroscience and Deep Learning DLSS16 2017-W1
3 2017, Aug, 29 Reinforcement I - Pineau - RL Basic Concepts 2017-W2
4 2017, Aug, 31 Generative I - GAN tutorial by Ian Goodfellow 2017-W2
5 2017, Sep, 5 Foundations III - Investigating Behaviors of DNN 2017-W3
6 2017, Sep, 7 Foundations IV - Investigating Behaviors of DNN 2017-W3
7 2017, Sep, 12 Foundations V - More about Behaviors of DNN 2017-W4
8 2017, Sep, 14 Foundations VI - More about Behaviors of DNN 2017-W4
9 2017, Sep, 19 Structure I - Varying DNN structures 2017-W5
10 2017, Sep, 21 Structure II - DNN with Varying Structures 2017-W5
11 2017, Sep, 26 Structure III - DNN with Attention 2017-W6
12 2017, Sep, 28 Structure IV - DNN with Attention 2 2017-W6
13 2017, Oct, 3 Structure V - DNN with Attention 3 2017-W7
14 2017, Oct, 5 Structure VI - DNN with Adaptive Structures 2017-W7
15 2017, Oct, 10 Reliable Applications I - Understanding 2017-W8
16 2017, Oct, 11 Reliable Applications V - Understanding2 2017-W10
17 2017, Oct, 12 Reliable Applications II - Data privacy 2017-W8
18 2017, Oct, 17 Reliable Applications III - Interesting Tasks 2017-W9
19 2017, Oct, 23 Reliable Applications IV - Robustness 2017-W9
20 2017, Oct, 26 Reliable Applications VI - Robustness2 2017-W10
21 2017, Oct, 31 Optimization I - Understanding DNN Optimization 2017-W11
22 2017, Nov, 2 Optimization II - DNN for Optimization 2017-W11
23 2017, Nov, 7 Optimization III - Optimization for DNN 2017-W12
24 2017, Nov, 9 Optimization IV - change DNN architecture for Optimization 2017-W12
25 2017, Nov, 14 Generative II - Deep Generative Models 2017-W13
26 2017, Nov, 16 Generative III - GAN training 2017-W13
27 2017, Nov, 21 RL II - Basic tutorial RLSS17 2017-W14
28 2017, Nov, 28 RL III - Basic tutorial RLSS17 (2) 2017-W14
29 2017, Nov, 30 RL IV - RL with varying structures 2017-W15

Foundations I -Andrew Ng - Nuts and Bolts of Applying Deep Learning

0Basics bias-variance
Presenter Papers Paper URL Our Slides
NIPS16 Andrew Ng - Nuts and Bolts of Applying Deep Learning: 1 video    
DLSS17 Doina Precup - Machine Learning - Bayesian Views (56:50m to 1:04:45 slides) video + slide    

bias variance both DNNflow

  1. Andrew Ng - Nuts and Bolts of Applying Deep Learning: 2016 + a few figures about Bias/variance trade-offs  

Foundations II - Ganguli - Theoretical Neuroscience and Deep Learning DLSS16

1Foundations neuroscience visualizing brain

Ganguli - Theoretical Neuroscience and Deep Learning

Presenter Papers Paper URL Our Slides
DLSS16 video    
DLSS17 video + slide    
DLSS17 Deep learning in the brain DLSS17 + Video  

Reinforcement I - Pineau - RL Basic Concepts

6Reinforcement 0Basics RL

Pineau - RL Basic Concepts

Presenter Papers Paper URL Our Slides
DLSS16 video    
RLSS17 slideRaw + video+ slide    

Generative I - GAN tutorial by Ian Goodfellow

5Generative 0Basics generative GAN
Presenter Papers Paper URL Our Slides
NIPS 2016 ganerative adversarial network tutorial (NIPS 2016) paper + video + code  
DLSS 2017 Generative Models I - DLSS 2017 slideraw + video + slide  

Foundations III - Investigating Behaviors of DNN

1Foundations understanding black-box generalization Expressive
Presenter Papers Paper URL Our Slides
Rita On the Expressive Power of Deep Neural Networks 1 PDF PDF
Arshdeep Understanding deep learning requires rethinking generalization, ICLR17 2 PDF PDF
Tianlu On Large-Batch Training for Deep Learning: Generalization Gap and Sharp Minima, ICLR17 3 PDF PDF
  1. On the Expressive Power of Deep Neural Networks / ICML 2017/ We propose a new approach to the problem of neural network expressivity, which seeks to characterize how structural properties of a neural network family affect the functions it is able to compute. Our approach is based on an interrelated set of measures of expressivity, unified by the novel notion of trajectory length, which measures how the output of a network changes as the input sweeps along a one-dimensional path. Our findings can be summarized as follows: (1) The complexity of the computed function grows exponentially with depth. (2) All weights are not equal: trained networks are more sensitive to their lower (initial) layer weights. (3) Regularizing on trajectory length (trajectory regularization) is a simpler alternative to batch normalization, with the same performance.  

  2. Understanding deep learning requires rethinking generalization, ICLR17 / Chiyuan Zhang, Samy Bengio, Moritz Hardt, Benjamin Recht, Oriol Vinyals (Submitted on 10 Nov 2016 (v1), last revised 26 Feb 2017 (this version, v2)) Despite their massive size, successful deep artificial neural networks can exhibit a remarkably small difference between training and test performance. Conventional wisdom attributes small generalization error either to properties of the model family, or to the regularization techniques used during training. Through extensive systematic experiments, we show how these traditional approaches fail to explain why large neural networks generalize well in practice. Specifically, our experiments establish that state-of-the-art convolutional networks for image classification trained with stochastic gradient methods easily fit a random labeling of the training data. This phenomenon is qualitatively unaffected by explicit regularization, and occurs even if we replace the true images by completely unstructured random noise. We corroborate these experimental findings with a theoretical construction showing that simple depth two neural networks already have perfect finite sample expressivity as soon as the number of parameters exceeds the number of data points as it usually does in practice. We interpret our experimental findings by comparison with traditional models.  

  3. On Large-Batch Training for Deep Learning: Generalization Gap and Sharp Minima, ICLR17 / Nitish Shirish Keskar, Dheevatsa Mudigere, Jorge Nocedal, Mikhail Smelyanskiy, Ping Tak Peter Tang/ The stochastic gradient descent (SGD) method and its variants are algorithms of choice for many Deep Learning tasks. These methods operate in a small-batch regime wherein a fraction of the training data, say 32-512 data points, is sampled to compute an approximation to the gradient. It has been observed in practice that when using a larger batch there is a degradation in the quality of the model, as measured by its ability to generalize. We investigate the cause for this generalization drop in the large-batch regime and present numerical evidence that supports the view that large-batch methods tend to converge to sharp minimizers of the training and testing functions - and as is well known, sharp minima lead to poorer generalization. In contrast, small-batch methods consistently converge to flat minimizers, and our experiments support a commonly held view that this is due to the inherent noise in the gradient estimation. We discuss several strategies to attempt to help large-batch methods eliminate this generalization gap.  

Foundations IV - Investigating Behaviors of DNN

1Foundations understanding black-box Parsimonious Associative memory
Presenter Papers Paper URL Our Slides
Beilun Learning Deep Parsimonious Representations, NIPS16 1 PDF PDF
Jack Dense Associative Memory for Pattern Recognition, NIPS16 2 PDF + video PDF
  1. Learning Deep Parsimonious Representations, NIPS16 / from Raquel Urtasun et al. / In this paper we aim at facilitating generalization for deep networks while supporting interpretability of the learned representations. Towards this goal, we propose a clustering based regularization that encourages parsimonious representations. Our k-means style objective is easy to optimize and flexible, supporting various forms of clustering, such as sample clustering, spatial clustering, as well as co-clustering. We demonstrate the effectiveness of our approach on the tasks of unsupervised learning, classification, fine grained categorization, and zero-shot learning.  

  2. Dense Associative Memory for Pattern Recognition, NIPS16 / from Hopfield/ A model of associative memory is studied, which stores and reliably retrieves many more patterns than the number of neurons in the network. We propose a simple duality between this dense associative memory and neural networks commonly used in deep learning. On the associative memory side of this duality, a family of models that smoothly interpolates between two limiting cases can be constructed. One limit is referred to as the feature-matching mode of pattern recognition, and the other one as the prototype regime. On the deep learning side of the duality, this family corresponds to feedforward neural networks with one hidden layer and various activation functions, which transmit the activities of the visible neurons to the hidden layer. This family of activation functions includes logistics, rectified linear units, and rectified polynomials of higher degrees. The proposed duality makes it possible to apply energy-based intuition from associative memory to analyze computational properties of neural networks with unusual activation functions - the higher rectified polynomials which until now have not been used in deep learning. The utility of the dense memories is illustrated for two test cases: the logical gate XOR and the recognition of handwritten digits from the MNIST data set.  

Foundations V - More about Behaviors of DNN

1Foundations understanding black-box Memorization InfoMax Expressive
Presenter Papers Paper URL Our Slides
Ceyer A Closer Look at Memorization in Deep Networks, ICML17 1 PDF PDF
  On the Expressive Efficiency of Overlapping Architectures of Deep Learning 2 DLSSpdf + video  
Mutual Information Opening the Black Box of Deep Neural Networks via Information 3 URL + video  
ChaoJiang Toward Deeper Understanding of Neural Networks: The Power of Initialization and a Dual View on Expressivity, NIPS16 PDF PDF
  1. A Closer Look at Memorization in Deep Networks, ICML17/ from etc Aaron Courville, Yoshua Bengio, Simon Lacoste-Julien / We examine the role of memorization in deep learning, drawing connections to capacity, generalization, and adversarial robustness. While deep networks are capable of memorizing noise data, our results suggest that they tend to prioritize learning simple patterns first. In our experiments, we expose qualitative differences in gradient-based optimization of deep neural networks (DNNs) on noise vs. real data. We also demonstrate that for appropriately tuned explicit regularization (e.g., dropout) we can degrade DNN training performance on noise datasets without compromising generalization on real data. Our analysis suggests that the notions of effective capacity which are dataset independent are unlikely to explain the generalization performance of deep networks when trained with gradient based methods because training data itself plays an important role in determining the degree of memorization.  

  2. On the Expressive Efficiency of Overlapping Architectures of Deep Learning / ICLR 2018/ Expressive efficiency refers to the relation between two architectures A and B, whereby any function realized by B could be replicated by A, but there exists functions realized by A, which cannot be replicated by B unless its size grows significantly larger. For example, it is known that deep networks are exponentially efficient with respect to shallow networks, in the sense that a shallow network must grow exponentially large in order to approximate the functions represented by a deep network of polynomial size. In this work, we extend the study of expressive efficiency to the attribute of network connectivity and in particular to the effect of “overlaps” in the convolutional process, i.e., when the stride of the convolution is smaller than its filter size (receptive field). To theoretically analyze this aspect of network’s design, we focus on a well-established surrogate for ConvNets called Convolutional Arithmetic Circuits (ConvACs), and then demonstrate empirically that our results hold for standard ConvNets as well. Specifically, our analysis shows that having overlapping local receptive fields, and more broadly denser connectivity, results in an exponential increase in the expressive capacity of neural networks. Moreover, while denser connectivity can increase the expressive capacity, we show that the most common types of modern architectures already exhibit exponential increase in expressivity, without relying on fully-connected layers.  

  3. Despite their great success, there is still no comprehensive theoretical understanding of learning with Deep Neural Networks (DNNs) or their inner organization. Previous work proposed to analyze DNNs in the \textit{Information Plane}; i.e., the plane of the Mutual Information values that each layer preserves on the input and output variables. They suggested that the goal of the network is to optimize the Information Bottleneck (IB) tradeoff between compression and prediction, successively, for each layer. In this work we follow up on this idea and demonstrate the effectiveness of the Information-Plane visualization of DNNs. Our main results are: (i) most of the training epochs in standard DL are spent on {\emph compression} of the input to efficient representation and not on fitting the training labels. (ii) The representation compression phase begins when the training errors becomes small and the Stochastic Gradient Decent (SGD) epochs change from a fast drift to smaller training error into a stochastic relaxation, or random diffusion, constrained by the training error value. (iii) The converged layers lie on or very close to the Information Bottleneck (IB) theoretical bound, and the maps from the input to any hidden layer and from this hidden layer to the output satisfy the IB self-consistent equations. This generalization through noise mechanism is unique to Deep Neural Networks and absent in one layer networks. (iv) The training time is dramatically reduced when adding more hidden layers. Thus the main advantage of the hidden layers is computational. This can be explained by the reduced relaxation time, as this it scales super-linearly (exponentially for simple diffusion) with the information compression from the previous layer.  

Foundations VI - More about Behaviors of DNN

1Foundations understanding black-box Expressive generalization
Presenter Papers Paper URL Our Slides
SE Equivariance Through Parameter-Sharing, ICML17 1 PDF  
SE Why Deep Neural Networks for Function Approximation?, ICLR17 2 PDF  
SE Geometry of Neural Network Loss Surfaces via Random Matrix Theory, 3ICML17 PDF  
  Sharp Minima Can Generalize For Deep Nets, ICML17 4 PDF  
  1. Sharp Minima Can Generalize For Deep Nets, ICML17 / Laurent Dinh, Razvan Pascanu, Samy Bengio, Yoshua Bengio / Despite their overwhelming capacity to overfit, deep learning architectures tend to generalize relatively well to unseen data, allowing them to be deployed in practice. However, explaining why this is the case is still an open area of research. One standing hypothesis that is gaining popularity, e.g. Hochreiter & Schmidhuber (1997); Keskar et al. (2017), is that the flatness of minima of the loss function found by stochastic gradient based methods results in good generalization. This paper argues that most notions of flatness are problematic for deep models and can not be directly applied to explain generalization. Specifically, when focusing on deep networks with rectifier units, we can exploit the particular geometry of parameter space induced by the inherent symmetries that these architectures exhibit to build equivalent models corresponding to arbitrarily sharper minima. Furthermore, if we allow to reparametrize a function, the geometry of its parameters can change drastically without affecting its generalization properties.  

  2. Measuring Sample Quality with Kernels, NIPS15/ To improve the efficiency of Monte Carlo estimation, practitioners are turning to biased Markov chain Monte Carlo procedures that trade off asymptotic exactness for computational speed. The reasoning is sound: a reduction in variance due to more rapid sampling can outweigh the bias introduced. However, the inexactness creates new challenges for sampler and parameter selection, since standard measures of sample quality like effective sample size do not account for asymptotic bias. To address these challenges, we introduce a new computable quality measure based on Stein’s method that quantifies the maximum discrepancy between sample and target expectations over a large class of test functions. We use our tool to compare exact, biased, and deterministic sample sequences and illustrate applications to hyperparameter selection, convergence rate assessment, and quantifying bias-variance tradeoffs in posterior inference.  

  3. Equivariance Through Parameter-Sharing, ICML17/ We propose to study equivariance in deep neural networks through parameter symmetries. In particular, given a group G that acts discretely on the input and output of a standard neural network layer ϕW:ℜM→ℜN, we show that ϕW is equivariant with respect to G-action iff G explains the symmetries of the network parameters W. Inspired by this observation, we then propose two parameter-sharing schemes to induce the desirable symmetry on W. Our procedures for tying the parameters achieve G-equivariance and, under some conditions on the action of , they guarantee sensitivity to all other permutation groups outside.  

  4. Geometry of Neural Network Loss Surfaces via Random Matrix Theory, ICML17 / Understanding the geometry of neural network loss surfaces is important for the development of improved optimization algorithms and for building a theoretical understanding of why deep learning works. In this paper, we study the geometry in terms of the distribution of eigenvalues of the Hessian matrix at critical points of varying energy. We introduce an analytical framework and a set of tools from random matrix theory that allow us to compute an approximation of this distribution under a set of simplifying assumptions. The shape of the spectrum depends strongly on the energy and another key parameter, ϕ, which measures the ratio of parameters to data points. Our analysis predicts and numerical simulations support that for critical points of small index, the number of negative eigenvalues scales like the 3/2 power of the energy. We leave as an open problem an explanation for our observation that, in the context of a certain memorization task, the energy of minimizers is well-approximated by the function 1/2(1−ϕ)^2.  

Structure I - Varying DNN structures

2Architecture 8Scalable dialog QA nonparametric structured sparsity
Presenter Papers Paper URL Our Slides
Jack Learning End-to-End Goal-Oriented Dialog, ICLR17 1 PDF PDF
Bargav Nonparametric Neural Networks, ICLR17 2 PDF PDF
Bargav Learning Structured Sparsity in Deep Neural Networks, NIPS16 3 PDF PDF
Arshdeep Learning the Number of Neurons in Deep Networks, NIPS16 4 PDF PDF
  1. Learning End-to-End Goal-Oriented Dialog, ICLR17 / Antoine Bordes, Y-Lan Boureau, Jason Weston/ Traditional dialog systems used in goal-oriented applications require a lot of domain-specific handcrafting, which hinders scaling up to new domains. End-to-end dialog systems, in which all components are trained from the dialogs themselves, escape this limitation. But the encouraging success recently obtained in chit-chat dialog may not carry over to goal-oriented settings. This paper proposes a testbed to break down the strengths and shortcomings of end-to-end dialog systems in goal-oriented applications. Set in the context of restaurant reservation, our tasks require manipulating sentences and symbols, so as to properly conduct conversations, issue API calls and use the outputs of such calls. We show that an end-to-end dialog system based on Memory Networks can reach promising, yet imperfect, performance and learn to perform non-trivial operations. We confirm those results by comparing our system to a hand-crafted slot-filling baseline on data from the second Dialog State Tracking Challenge (Henderson et al., 2014a). We show similar result patterns on data extracted from an online concierge service.  

  2. Nonparametric Neural Networks, ICLR17 / George Philipp, Jaime G. Carbonell/ Automatically determining the optimal size of a neural network for a given task without prior information currently requires an expensive global search and training many networks from scratch. In this paper, we address the problem of automatically finding a good network size during a single training cycle. We introduce nonparametric neural networks, a non-probabilistic framework for conducting optimization over all possible network sizes and prove its soundness when network growth is limited via an L_p penalty. We train networks under this framework by continuously adding new units while eliminating redundant units via an L_2 penalty. We employ a novel optimization algorithm, which we term adaptive radial-angular gradient descent or AdaRad, and obtain promising results.  

  3. Learning Structured Sparsity in Deep Neural Networks, NIPS16/ High demand for computation resources severely hinders deployment of large-scale Deep Neural Networks (DNN) in resource constrained devices. In this work, we propose a Structured Sparsity Learning (SSL) method to regularize the structures (i.e., filters, channels, filter shapes, and layer depth) of DNNs. SSL can: (1) learn a compact structure from a bigger DNN to reduce computation cost; (2) obtain a hardware-friendly structured sparsity of DNN to efficiently accelerate the DNN’s evaluation. Experimental results show that SSL achieves on average 5.1 × and 3.1 × speedups of convolutional layer computation of AlexNet against CPU and GPU, respectively, with off-the-shelf libraries. These speedups are about twice speedups of non-structured sparsity; (3) regularize the DNN structure to improve classification accuracy. The results show that for CIFAR-10, regularization on layer depth reduces a 20-layer Deep Residual Network (ResNet) to 18 layers while improves the accuracy from 91.25% to 92.60%, which is still higher than that of original ResNet with 32 layers. For AlexNet, SSL reduces the error by ~ 1%.  

  4. Learning the Number of Neurons in Deep Networks, NIPS16 / Nowadays, the number of layers and of neurons in each layer of a deep network are typically set manually. While very deep and wide networks have proven effective in general, they come at a high memory and computation cost, thus making them impractical for constrained platforms. These networks, however, are known to have many redundant parameters, and could thus, in principle, be replaced by more compact architectures. In this paper, we introduce an approach to automatically determining the number of neurons in each layer of a deep network during learning. To this end, we propose to make use of a group sparsity regularizer on the parameters of the network, where each group is defined to act on a single neuron. Starting from an overcomplete network, we show that our approach can reduce the number of parameters by up to 80\% while retaining or even improving the network accuracy.  

Structure II - DNN with Varying Structures

2Architecture 8Scalable sparsity blocking nonparametric structured QA Interpretable
Presenter Papers Paper URL Our Slides
Shijia Outrageously Large Neural Networks: The Sparsely-Gated Mixture-of-Experts Layer, (Dean), ICLR17 1 PDF PDF
Ceyer Sequence Modeling via Segmentations, ICML17 2 PDF PDF
Arshdeep Input Switched Affine Networks: An RNN Architecture Designed for Interpretability, ICML17 3 PDF PDF


  1. Outrageously Large Neural Networks: The Sparsely-Gated Mixture-of-Experts Layer, (Dean), ICLR17/ The capacity of a neural network to absorb information is limited by its number of parameters. Conditional computation, where parts of the network are active on a per-example basis, has been proposed in theory as a way of dramatically increasing model capacity without a proportional increase in computation. In practice, however, there are significant algorithmic and performance challenges. In this work, we address these challenges and finally realize the promise of conditional computation, achieving greater than 1000x improvements in model capacity with only minor losses in computational efficiency on modern GPU clusters. We introduce a Sparsely-Gated Mixture-of-Experts layer (MoE), consisting of up to thousands of feed-forward sub-networks. A trainable gating network determines a sparse combination of these experts to use for each example. We apply the MoE to the tasks of language modeling and machine translation, where model capacity is critical for absorbing the vast quantities of knowledge available in the training corpora. We present model architectures in which a MoE with up to 137 billion parameters is applied convolutionally between stacked LSTM layers. On large language modeling and machine translation benchmarks, these models achieve significantly better results than state-of-the-art at lower computational cost. 

  2. Sequence Modeling via Segmentations, ICML17/ Segmental structure is a common pattern in many types of sequences such as phrases in human languages. In this paper, we present a probabilistic model for sequences via their segmentations. The probability of a segmented sequence is calculated as the product of the probabilities of all its segments, where each segment is modeled using existing tools such as recurrent neural networks. Since the segmentation of a sequence is usually unknown in advance, we sum over all valid segmentations to obtain the final probability for the sequence. An efficient dynamic programming algorithm is developed for forward and backward computations without resorting to any approximation. We demonstrate our approach on text segmentation and speech recognition tasks. In addition to quantitative results, we also show that our approach can discover meaningful segments in their respective application contexts.  

  3. Input Switched Affine Networks: An RNN Architecture Designed for Interpretability, ICML17/ There exist many problem domains where the interpretability of neural network models is essential for deployment. Here we introduce a recurrent architecture composed of input-switched affine transformations - in other words an RNN without any explicit nonlinearities, but with input-dependent recurrent weights. This simple form allows the RNN to be analyzed via straightforward linear methods: we can exactly characterize the linear contribution of each input to the model predictions; we can use a change-of-basis to disentangle input, output, and computational hidden unit subspaces; we can fully reverse-engineer the architecture’s solution to a simple task. Despite this ease of interpretation, the input switched affine network achieves reasonable performance on a text modeling tasks, and allows greater computational efficiency than networks with standard nonlinearities.  

Structure III - DNN with Attention

2Architecture attention transfer-learning dynamic structured QA relational
Presenter Papers Paper URL Our Slides
Rita Paying More Attention to Attention: Improving the Performance of Convolutional Neural Networks via Attention Transfer, ICLR17 1 PDF PDF
Tianlu Dynamic Coattention Networks For Question Answering, ICLR17 2 PDF + code PDF
ChaoJiang Structured Attention Networks, ICLR17 3 PDF + code PDF
  1. Paying More Attention to Attention: Improving the Performance of Convolutional Neural Networks via Attention Transfer, ICLR17 / Attention plays a critical role in human visual experience. Furthermore, it has recently been demonstrated that attention can also play an important role in the context of applying artificial neural networks to a variety of tasks from fields such as computer vision and NLP. In this work we show that, by properly defining attention for convolutional neural networks, we can actually use this type of information in order to significantly improve the performance of a student CNN network by forcing it to mimic the attention maps of a powerful teacher network. To that end, we propose several novel methods of transferring attention, showing consistent improvement across a variety of datasets and convolutional neural network architectures. Code and models for our experiments are available at this https URL.  

  2. Dynamic Coattention Networks For Question Answering, ICLR17 / Caiming Xiong, Victor Zhong, Richard Socher/ Several deep learning models have been proposed for question answering. However, due to their single-pass nature, they have no way to recover from local maxima corresponding to incorrect answers. To address this problem, we introduce the Dynamic Coattention Network (DCN) for question answering. The DCN first fuses co-dependent representations of the question and the document in order to focus on relevant parts of both. Then a dynamic pointing decoder iterates over potential answer spans. This iterative procedure enables the model to recover from initial local maxima corresponding to incorrect answers. On the Stanford question answering dataset, a single DCN model improves the previous state of the art from 71.0% F1 to 75.9%, while a DCN ensemble obtains 80.4% F1.  

  3. Structured Attention Networks, ICLR17 / Attention networks have proven to be an effective approach for embedding categorical inference within a deep neural network. However, for many tasks we may want to model richer structural dependencies without abandoning end-to-end training. In this work, we experiment with incorporating richer structural distributions, encoded using graphical models, within deep networks. We show that these structured attention networks are simple extensions of the basic attention procedure, and that they allow for extending attention beyond the standard soft-selection approach, such as attending to partial segmentations or to subtrees. We experiment with two different classes of structured attention networks: a linear-chain conditional random field and a graph-based parsing model, and describe how these models can be practically implemented as neural network layers. Experiments show that this approach is effective for incorporating structural biases, and structured attention networks outperform baseline attention models on a variety of synthetic and real tasks: tree transduction, neural machine translation, question answering, and natural language inference. We further find that models trained in this way learn interesting unsupervised hidden representations that generalize simple attention.  

Structure IV - DNN with Attention 2

2Architecture attention transfer-learning relational generative memory Infomax
Presenter Papers Paper URL Our Slides
Jack Attend, Adapt and Transfer: Attentive Deep Architecture for Adaptive Transfer from multiple sources in the same domain, ICLR17 1 PDF PDF
Arshdeep Bidirectional Attention Flow for Machine Comprehension, ICLR17 2 PDF + code PDF
Ceyer Image-to-Markup Generation with Coarse-to-Fine Attention, ICML17 PDF + code PDF
ChaoJiang Can Active Memory Replace Attention? ; Samy Bengio, NIPS16 3 PDF PDF
  An Information-Theoretic Framework for Fast and Robust Unsupervised Learning via Neural Population Infomax, ICLR17 PDF  
  1. Bidirectional Attention Flow for Machine Comprehension, ICLR17 /Machine comprehension (MC), answering a query about a given context paragraph, requires modeling complex interactions between the context and the query. Recently, attention mechanisms have been successfully extended to MC. Typically these methods use attention to focus on a small portion of the context and summarize it with a fixed-size vector, couple attentions temporally, and/or often form a uni-directional attention. In this paper we introduce the Bi-Directional Attention Flow (BIDAF) network, a multi-stage hierarchical process that represents the context at different levels of granularity and uses bi-directional attention flow mechanism to obtain a query-aware context representation without early summarization. Our experimental evaluations show that our model achieves the state-of-the-art results in Stanford Question Answering Dataset (SQuAD) and CNN/DailyMail cloze test.  

  2. Image-to-Markup Generation with Coarse-to-Fine Attention, ICML17/ We present a neural encoder-decoder model to convert images into presentational markup based on a scalable coarse-to-fine attention mechanism. Our method is evaluated in the context of image-to-LaTeX generation, and we introduce a new dataset of real-world rendered mathematical expressions paired with LaTeX markup. We show that unlike neural OCR techniques using CTC-based models, attention-based approaches can tackle this non-standard OCR task. Our approach outperforms classical mathematical OCR systems by a large margin on in-domain rendered data, and, with pretraining, also performs well on out-of-domain handwritten data. To reduce the inference complexity associated with the attention-based approaches, we introduce a new coarse-to-fine attention layer that selects a support region before applying attention.  

  3. Can Active Memory Replace Attention? ; Samy Bengio, NIPS16/ Several mechanisms to focus attention of a neural network on selected parts of its input or memory have been used successfully in deep learning models in recent years. Attention has improved image classification, image captioning, speech recognition, generative models, and learning algorithmic tasks, but it had probably the largest impact on neural machine translation. Recently, similar improvements have been obtained using alternative mechanisms that do not focus on a single part of a memory but operate on all of it in parallel, in a uniform way. Such mechanism, which we call active memory, improved over attention in algorithmic tasks, image processing, and in generative modelling. So far, however, active memory has not improved over attention for most natural language processing tasks, in particular for machine translation. We analyze this shortcoming in this paper and propose an extended model of active memory that matches existing attention models on neural machine translation and generalizes better to longer sentences. We investigate this model and explain why previous active memory models did not succeed. Finally, we discuss when active memory brings most benefits and where attention can be a better choice.  

Structure V - DNN with Attention 3

2Architecture dynamic QA memory
Presenter Papers Paper URL Our Slides
Tianlu Ask Me Anything: Dynamic Memory Networks for Natural Language Processing, ICML17 1 PDF + code PDF
Jack Reasoning with Memory Augmented Neural Networks for Language Comprehension, ICLR17 2 PDF PDF
Xueying State-Frequency Memory Recurrent Neural Networks, ICML17 3 PDF PDF
  1. Ask Me Anything: Dynamic Memory Networks for Natural Language Processing, ICML16 / Most tasks in natural language processing can be cast into question answering (QA) problems over language input. We introduce the dynamic memory network (DMN), a neural network architecture which processes input sequences and questions, forms episodic memories, and generates relevant answers. Questions trigger an iterative attention process which allows the model to condition its attention on the inputs and the result of previous iterations. These results are then reasoned over in a hierarchical recurrent sequence model to generate answers. The DMN can be trained end-to-end and obtains state-of-the-art results on several types of tasks and datasets: question answering (Facebook’s bAbI dataset), text classification for sentiment analysis (Stanford Sentiment Treebank) and sequence modeling for part-of-speech tagging (WSJ-PTB). The training for these different tasks relies exclusively on trained word vector representations and input-question-answer triplets. (high citation)  

  2. Reasoning with Memory Augmented Neural Networks for Language Comprehension, ICLR17 / Hypothesis testing is an important cognitive process that supports human reasoning. In this paper, we introduce a computational hypothesis testing approach based on memory augmented neural networks. Our approach involves a hypothesis testing loop that reconsiders and progressively refines a previously formed hypothesis in order to generate new hypotheses to test. We apply the proposed approach to language comprehension task by using Neural Semantic Encoders (NSE). Our NSE models achieve the state-of-the-art results showing an absolute improvement of 1.2% to 2.6% accuracy over previous results obtained by single and ensemble systems on standard machine comprehension benchmarks such as the Children’s Book Test (CBT) and Who-Did-What (WDW) news article datasets.  

  3. State-Frequency Memory Recurrent Neural Networks, ICML17/ Modeling temporal sequences plays a fundamental role in various modern applications and has drawn more and more attentions in the machine learning community. Among those efforts on improving the capability to represent temporal data, the Long Short-Term Memory (LSTM) has achieved great success in many areas. Although the LSTM can capture long-range dependency in the time domain, it does not explicitly model the pattern occurrences in the frequency domain that plays an important role in tracking and predicting data points over various time cycles. We propose the State-Frequency Memory (SFM), a novel recurrent architecture that allows to separate dynamic patterns across different frequency components and their impacts on modeling the temporal contexts of input sequences. By jointly decomposing memorized dynamics into state-frequency components, the SFM is able to offer a fine-grained analysis of temporal sequences by capturing the dependency of uncovered patterns in both time and frequency domains. Evaluations on several temporal modeling tasks demonstrate the SFM can yield competitive performances, in particular as compared with the state-of-the-art LSTM models.  

Structure VI - DNN with Adaptive Structures

2Architecture 8Scalable dynamic Architecture Search structured
Presenter Papers Paper URL Our Slides
Anant AdaNet: Adaptive Structural Learning of Artificial Neural Networks, ICML17 1 PDF PDF
Shijia SplitNet: Learning to Semantically Split Deep Networks for Parameter Reduction and Model Parallelization, ICML17 2 PDF PDF
Jack Proximal Deep Structured Models, NIPS16 3 PDF PDF
  Optimal Architectures in a Solvable Model of Deep Networks, NIPS16 4 PDF  
Tianlu Large-Scale Evolution of Image Classifiers, ICML17 5 PDF PDF
  1. Large-Scale Evolution of Image Classifiers, ICML17/ Esteban Real, Sherry Moore, Andrew Selle, Saurabh Saxena, Yutaka Leon Suematsu, Jie Tan, Quoc Le, Alex Kurakin / Neural networks have proven effective at solving difficult problems but designing their architectures can be challenging, even for image classification problems alone. Our goal is to minimize human participation, so we employ evolutionary algorithms to discover such networks automatically. Despite significant computational requirements, we show that it is now possible to evolve models with accuracies within the range of those published in the last year. Specifically, we employ simple evolutionary techniques at unprecedented scales to discover models for the CIFAR-10 and CIFAR-100 datasets, starting from trivial initial conditions and reaching accuracies of 94.6% (95.6% for ensemble) and 77.0%, respectively. To do this, we use novel and intuitive mutation operators that navigate large search spaces; we stress that no human participation is required once evolution starts and that the output is a fully-trained model. Throughout this work, we place special emphasis on the repeatability of results, the variability in the outcomes and the computational requirements.  

  2. AdaNet: Adaptive Structural Learning of Artificial Neural Networks, ICML17 / Corinna Cortes, et al. We present new algorithms for adaptively learning artificial neural networks. Our algorithms (AdaNet) adaptively learn both the structure of the network and its weights. They are based on a solid theoretical analysis, including data-dependent generalization guarantees that we prove and discuss in detail. We report the results of large-scale experiments with one of our algorithms on several binary classification tasks extracted from the CIFAR-10 dataset. The results demonstrate that our algorithm can automatically learn network structures with very competitive performance accuracies when compared with those achieved for neural networks found by standard approaches.  

  3. SplitNet: Learning to Semantically Split Deep Networks for Parameter Reduction and Model Parallelization, ICML17 / We propose a novel deep neural network that is both lightweight and effectively structured for model parallelization. Our network, which we name as SplitNet, automatically learns to split the network weights into either a set or a hierarchy of multiple groups that use disjoint sets of features, by learning both the class-to-group and feature-to-group assignment matrices along with the network weights. This produces a tree-structured network that involves no connection between branched subtrees of semantically disparate class groups. SplitNet thus greatly reduces the number of parameters and requires significantly less computations, and is also embarrassingly model parallelizable at test time, since the network evaluation for each subnetwork is completely independent except for the shared lower layer weights that can be duplicated over multiple processors. We validate our method with two deep network models (ResNet and AlexNet) on two different datasets (CIFAR-100 and ILSVRC 2012) for image classification, on which our method obtains networks with significantly reduced number of parameters while achieving comparable or superior classification accuracies over original full deep networks, and accelerated test speed with multiple GPUs.  

  4. Proximal Deep Structured Models, NIPS16 / Raquel Urtasun et al. Many problems in real-world applications involve predicting continuous-valued random variables that are statistically related. In this paper, we propose a powerful deep structured model that is able to learn complex non-linear functions which encode the dependencies between continuous output variables. We show that inference in our model using proximal methods can be efficiently solved as a feedfoward pass of a special type of deep recurrent neural network. We demonstrate the effectiveness of our approach in the tasks of image denoising, depth refinement and optical flow estimation.  

  5. Optimal architectures in a solvable model of Deep networks, NIPS16/ Deep neural networks have received a considerable attention due to the success of their training for real world machine learning applications. They are also of great interest to the understanding of sensory processing in cortical sensory hierarchies. The purpose of this work is to advance our theoretical understanding of the computational benefits of these architectures. Using a simple model of clustered noisy inputs and a simple learning rule, we provide analytically derived recursion relations describing the propagation of the signals along the deep network. By analysis of these equations, and defining performance measures, we show that these model networks have optimal depths. We further explore the dependence of the optimal architecture on the system parameters.  

Reliable Applications I - Understanding

3Reliable Interpretable Model-Criticism random sample-quality Difference-Analysis Attribution
Presenter Papers Paper URL Our Slides
Rita Learning Important Features Through Propagating Activation Differences, ICML17 1 PDF PDF
GaoJi Examples are not Enough, Learn to Criticize! Model Criticism for Interpretable Machine Learning, NIPS16 2 PDF PDF
Rita Learning Kernels with Random Features, Aman Sinha*; John Duchi, 3 PDF PDF
  Measuring Sample Quality with Kernels, NIPS16 [^4] PDF  
  1. Learning Kernels with Random Features / John Duchi NIPS2016/ Randomized features provide a computationally efficient way to approximate kernel machines in machine learning tasks. However, such methods require a user-defined kernel as input. We extend the randomized-feature approach to the task of learning a kernel (via its associated random features). Specifically, we present an efficient optimization problem that learns a kernel in a supervised manner. We prove the consistency of the estimated kernel as well as generalization bounds for the class of estimators induced by the optimized kernel, and we experimentally evaluate our technique on several datasets. Our approach is efficient and highly scalable, and we attain competitive results with a fraction of the training cost of other techniques.  

  2. Learning Important Features Through Propagating Activation Differences, ICML17 / DeepLIFE / The purported “black box”’ nature of neural networks is a barrier to adoption in applications where interpretability is essential. Here we present DeepLIFT (Deep Learning Important FeaTures), a method for decomposing the output prediction of a neural network on a specific input by backpropagating the contributions of all neurons in the network to every feature of the input. DeepLIFT compares the activation of each neuron to its ‘reference activation’ and assigns contribution scores according to the difference. By optionally giving separate consideration to positive and negative contributions, DeepLIFT can also reveal dependencies which are missed by other approaches. Scores can be computed efficiently in a single backward pass. We apply DeepLIFT to models trained on MNIST and simulated genomic data, and show significant advantages over gradient-based methods. A detailed video tutorial on the method is at this http URL and code is at this http URL.  

  3. Examples are not Enough, Learn to Criticize! Model Criticism for Interpretable Machine Learning, NIPS16 / Been Kim et al. / Example-based explanations are widely used in the effort to improve the interpretability of highly complex distributions. However, prototypes alone are rarely sufficient to represent the gist of the complexity. In order for users to construct better mental models and understand complex data distributions, we also need criticism to explain what are not captured by prototypes. Motivated by the Bayesian model criticism framework, we develop MMD-critic which efficiently learns prototypes and criticism, designed to aid human interpretability. A human subject pilot study shows that the MMD-critic selects prototypes and criticism that are useful to facilitate human understanding and reasoning. We also evaluate the prototypes selected by MMD-critic via a nearest prototype classifier, showing competitive performance compared to baselines.  

Reliable Applications V - Understanding2

3Reliable visualizing Difference-Analysis Attribution Composition
Presenter Papers Paper URL Our Slides
Rita Visualizing Deep Neural Network Decisions: Prediction Difference Analysis, ICLR17 1 PDF PDF
Arshdeep Axiomatic Attribution for Deep Networks, ICML17 2 PDF PDF
  The Robustness of Estimator Composition, NIPS16 PDF  
  1. Axiomatic Attribution for Deep Networks, ICML17 / Google / We study the problem of attributing the prediction of a deep network to its input features, a problem previously studied by several other works. We identify two fundamental axioms—Sensitivity and Implementation Invariance that attribution methods ought to satisfy. We show that they are not satisfied by most known attribution methods, which we consider to be a fundamental weakness of those methods. We use the axioms to guide the design of a new attribution method called Integrated Gradients. Our method requires no modification to the original network and is extremely simple to implement; it just needs a few calls to the standard gradient operator. We apply this method to a couple of image models, a couple of text models and a chemistry model, demonstrating its ability to debug networks, to extract rules from a network, and to enable users to engage with models better.  

  2. Toward Deeper Understanding of Neural Networks: The Power of Initialization and a Dual View on Expressivity, NIPS16 / GoogleBrain / We develop a general duality between neural networks and compositional kernel Hilbert spaces. We introduce the notion of a computation skeleton, an acyclic graph that succinctly describes both a family of neural networks and a kernel space. Random neural networks are generated from a skeleton through node replication followed by sampling from a normal distribution to assign weights. The kernel space consists of functions that arise by compositions, averaging, and non-linear transformations governed by the skeleton’s graph topology and activation functions. We prove that random networks induce representations which approximate the kernel space. In particular, it follows that random weight initialization often yields a favorable starting point for optimization despite the worst-case intractability of training neural networks.  

Reliable Applications II - Data privacy

3Reliable Semi-supervised Privacy Domain-adaptation
Presenter Papers Paper URL Our Slides
Xueying Semi-supervised Knowledge Transfer for Deep Learning from Private Training Data, ICLR17 1 PDF PDF
Bargav Deep Learning with Differential Privacy, CCS16 2 PDF + video PDF
Bargav Privacy-Preserving Deep Learning, CCS15 3 PDF PDF
Xueying Domain Separation Networks, NIPS16 4 PDF PDF
  1. Semi-supervised Knowledge Transfer for Deep Learning from Private Training Data, ICLR17 / Nicolas Papernot et al. Some machine learning applications involve training data that is sensitive, such as the medical histories of patients in a clinical trial. A model may inadvertently and implicitly store some of its training data; careful analysis of the model may therefore reveal sensitive information. To address this problem, we demonstrate a generally applicable approach to providing strong privacy guarantees for training data: Private Aggregation of Teacher Ensembles (PATE). The approach combines, in a black-box fashion, multiple models trained with disjoint datasets, such as records from different subsets of users. Because they rely directly on sensitive data, these models are not published, but instead used as “teachers” for a “student” model. The student learns to predict an output chosen by noisy voting among all of the teachers, and cannot directly access an individual teacher or the underlying data or parameters. The student’s privacy properties can be understood both intuitively (since no single teacher and thus no single dataset dictates the student’s training) and formally, in terms of differential privacy. These properties hold even if an adversary can not only query the student but also inspect its internal workings. Compared with previous work, the approach imposes only weak assumptions on how teachers are trained: it applies to any model, including non-convex models like DNNs. We achieve state-of-the-art privacy/utility trade-offs on MNIST and SVHN thanks to an improved privacy analysis and semi-supervised learning.  

  2. Domain Separation Networks, NIPS16 / The cost of large scale data collection and annotation often makes the application of machine learning algorithms to new tasks or datasets prohibitively expensive. One approach circumventing this cost is training models on synthetic data where annotations are provided automatically. Despite their appeal, such models often fail to generalize from synthetic to real images, necessitating domain adaptation algorithms to manipulate these models before they can be successfully applied. Existing approaches focus either on mapping representations from one domain to the other, or on learning to extract features that are invariant to the domain from which they were extracted. However, by focusing only on creating a mapping or shared representation between the two domains, they ignore the individual characteristics of each domain. We hypothesize that explicitly modeling what is unique to each domain can improve a model’s ability to extract domain-invariant features. Inspired by work on private-shared component analysis, we explicitly learn to extract image representations that are partitioned into two subspaces: one component which is private to each domain and one which is shared across domains. Our model is trained to not only perform the task we care about in the source domain, but also to use the partitioned representation to reconstruct the images from both domains. Our novel architecture results in a model that outperforms the state-of-the-art on a range of unsupervised domain adaptation scenarios and additionally produces visualizations of the private and shared representations enabling interpretation of the domain adaptation process.  

  3. Deep Learning with Differential Privacy, CCS 2016/ Machine learning techniques based on neural networks are achieving remarkable results in a wide variety of domains. Often, the training of models requires large, representative datasets, which may be crowdsourced and contain sensitive information. The models should not expose private information in these datasets. Addressing this goal, we develop new algorithmic techniques for learning and a refined analysis of privacy costs within the framework of differential privacy. Our implementation and experiments demonstrate that we can train deep neural networks with non-convex objectives, under a modest privacy budget, and at a manageable cost in software complexity, training efficiency, and model quality.  

  4. Privacy-Preserving Deep Learning, CCS15/ Deep learning based on artificial neural networks is a very popular approach to modeling, classifying, and recognizing complex data such as images, speech, and text. The unprecedented accuracy of deep learning methods has turned them into the foundation of new AI-based services on the Internet. Commercial companies that collect user data on a large scale have been the main beneficiaries of this trend since the success of deep learning techniques is directly proportional to the amount of data available for training. Massive data collection required for deep learning presents obvious privacy issues. Users’ personal, highly sensitive data such as photos and voice recordings is kept indefinitely by the companies that collect it. Users can neither delete it, nor restrict the purposes for which it is used. Furthermore, centrally kept data is subject to legal subpoenas and extra-judicial surveillance. Many data owners–for example, medical institutions that may want to apply deep learning methods to clinical records–are prevented by privacy and confidentiality concerns from sharing the data and thus benefitting from large-scale deep learning. In this paper, we design, implement, and evaluate a practical system that enables multiple parties to jointly learn an accurate neural-network model for a given objective without sharing their input datasets. We exploit the fact that the optimization algorithms used in modern deep learning, namely, those based on stochastic gradient descent, can be parallelized and executed asynchronously. Our system lets participants train independently on their own datasets and selectively share small subsets of their models’ key parameters during training. This offers an attractive point in the utility/privacy tradeoff space: participants preserve the privacy of their respective data while still benefitting from other participants’ models and thus boosting their learning accuracy beyond what is achievable solely on their own inputs. We demonstrate the accuracy of our privacy-preserving deep learning on benchmark datasets.  

Reliable Applications III - Interesting Tasks

3Reliable QA noise Neural-Programming Hierarchical
Presenter Papers Paper URL Our Slides
Jack Learning to Query, Reason, and Answer Questions On Ambiguous Texts, ICLR17 1 PDF PDF
Arshdeep Making Neural Programming Architectures Generalize via Recursion, ICLR17 2 PDF PDF
Xueying Towards Deep Interpretability (MUS-ROVER II): Learning Hierarchical Representations of Tonal Music, ICLR17 3 PDF PDF
  1. Making Neural Programming Architectures Generalize via Recursion, ICLR17 / Jonathon Cai, Richard Shin, Dawn Song/ Empirically, neural networks that attempt to learn programs from data have exhibited poor generalizability. Moreover, it has traditionally been difficult to reason about the behavior of these models beyond a certain level of input complexity. In order to address these issues, we propose augmenting neural architectures with a key abstraction: recursion. As an application, we implement recursion in the Neural Programmer-Interpreter framework on four tasks: grade-school addition, bubble sort, topological sort, and quicksort. We demonstrate superior generalizability and interpretability with small amounts of training data. Recursion divides the problem into smaller pieces and drastically reduces the domain of each neural network component, making it tractable to prove guarantees about the overall system’s behavior. Our experience suggests that in order for neural architectures to robustly learn program semantics, it is necessary to incorporate a concept like recursion.  

  2. Learning to Query, Reason, and Answer Questions On Ambiguous Texts, ICLR17/ A key goal of research in conversational systems is to train an interactive agent to help a user with a task. Human conversation, however, is notoriously incomplete, ambiguous, and full of extraneous detail. To operate effectively, the agent must not only understand what was explicitly conveyed but also be able to reason in the presence of missing or unclear information. When unable to resolve ambiguities on its own, the agent must be able to ask the user for the necessary clarifications and incorporate the response in its reasoning. Motivated by this problem we introduce QRAQ (“crack”; Query, Reason, and Answer Questions), a new synthetic domain, in which a User gives an Agent a short story and asks a challenge question. These problems are designed to test the reasoning and interaction capabilities of a learning-based Agent in a setting that requires multiple conversational turns. A good Agent should ask only non-deducible, relevant questions until it has enough information to correctly answer the User’s question. We use standard and improved reinforcement learning based memory-network architectures to solve QRAQ problems in the difficult setting where the reward signal only tells the Agent if its final answer to the challenge question is correct or not. To provide an upper-bound to the RL results we also train the same architectures using supervised information that tells the Agent during training which variables to query and the answer to the challenge question. We evaluate our architectures on four QRAQ dataset types, and scale the complexity for each along multiple dimensions.  

  3. Towards Deep Interpretability (MUS-ROVER II): Learning Hierarchical Representations of Tonal Music, ICLR17 / Music theory studies the regularity of patterns in music to capture concepts underlying music styles and composers’ decisions. This paper continues the study of building \emph{automatic theorists} (rovers) to learn and represent music concepts that lead to human interpretable knowledge and further lead to materials for educating people. Our previous work took a first step in algorithmic concept learning of tonal music, studying high-level representations (concepts) of symbolic music (scores) and extracting interpretable rules for composition. This paper further studies the representation \emph{hierarchy} through the learning process, and supports \emph{adaptive} 2D memory selection in the resulting language model. This leads to a deeper-level interpretability that expands from individual rules to a dynamic system of rules, making the entire rule learning process more cognitive. The outcome is a new rover, MUS-ROVER \RN{2}, trained on Bach’s chorales, which outputs customizable syllabi for learning compositional rules. We demonstrate comparable results to our music pedagogy, while also presenting the differences and variations. In addition, we point out the rover’s potential usages in style recognition and synthesis, as well as applications beyond music.  

Reliable Applications IV - Robustness

3Reliable Adversarial-Examples high-dimensional robustness
Presenter Papers Paper URL Our Slides
GaoJi Delving into Transferable Adversarial Examples and Black-box Attacks,ICLR17 1 pdf PDF
Shijia On Detecting Adversarial Perturbations, ICLR17 2 pdf PDF
Anant Parseval Networks: Improving Robustness to Adversarial Examples, ICML17 3 pdf PDF
Bargav Being Robust (in High Dimensions) Can Be Practical, ICML17 4 pdf PDF
  1. Delving into Transferable Adversarial Examples and Black-box Attacks,ICLR17 / Down Song et al, high cite / An intriguing property of deep neural networks is the existence of adversarial examples, which can transfer among different architectures. These transferable adversarial examples may severely hinder deep neural network-based applications. Previous works mostly study the transferability using small scale datasets. In this work, we are the first to conduct an extensive study of the transferability over large models and a large scale dataset, and we are also the first to study the transferability of targeted adversarial examples with their target labels. We study both non-targeted and targeted adversarial examples, and show that while transferable non-targeted adversarial examples are easy to find, targeted adversarial examples generated using existing approaches almost never transfer with their target labels. Therefore, we propose novel ensemble-based approaches to generating transferable adversarial examples. Using such approaches, we observe a large proportion of targeted adversarial examples that are able to transfer with their target labels for the first time. We also present some geometric studies to help understanding the transferable adversarial examples. Finally, we show that the adversarial examples generated using ensemble-based approaches can successfully attack, which is a black-box image classification system.  

  2. On Detecting Adversarial Perturbations, ICLR17 / Machine learning and deep learning in particular has advanced tremendously on perceptual tasks in recent years. However, it remains vulnerable against adversarial perturbations of the input that have been crafted specifically to fool the system while being quasi-imperceptible to a human. In this work, we propose to augment deep neural networks with a small “detector” subnetwork which is trained on the binary classification task of distinguishing genuine data from data containing adversarial perturbations. Our method is orthogonal to prior work on addressing adversarial perturbations, which has mostly focused on making the classification network itself more robust. We show empirically that adversarial perturbations can be detected surprisingly well even though they are quasi-imperceptible to humans. Moreover, while the detectors have been trained to detect only a specific adversary, they generalize to similar and weaker adversaries. In addition, we propose an adversarial attack that fools both the classifier and the detector and a novel training procedure for the detector that counteracts this attack.  

  3. Parseval Networks: Improving Robustness to Adversarial Examples, ICML17 / We introduce Parseval networks, a form of deep neural networks in which the Lipschitz constant of linear, convolutional and aggregation layers is constrained to be smaller than 1. Parseval networks are empirically and theoretically motivated by an analysis of the robustness of the predictions made by deep neural networks when their input is subject to an adversarial perturbation. The most important feature of Parseval networks is to maintain weight matrices of linear and convolutional layers to be (approximately) Parseval tight frames, which are extensions of orthogonal matrices to non-square matrices. We describe how these constraints can be maintained efficiently during SGD. We show that Parseval networks match the state-of-the-art in terms of accuracy on CIFAR-10/100 and Street View House Numbers (SVHN) while being more robust than their vanilla counterpart against adversarial examples. Incidentally, Parseval networks also tend to train faster and make a better usage of the full capacity of the networks.  

  4. Being Robust (in High Dimensions) Can Be Practical, ICML17/ Robust estimation is much more challenging in high dimensions than it is in one dimension: Most techniques either lead to intractable optimization problems or estimators that can tolerate only a tiny fraction of errors. Recent work in theoretical computer science has shown that, in appropriate distributional models, it is possible to robustly estimate the mean and covariance with polynomial time algorithms that can tolerate a constant fraction of corruptions, independent of the dimension. However, the sample and time complexity of these algorithms is prohibitively large for high-dimensional applications. In this work, we address both of these issues by establishing sample complexity bounds that are optimal, up to logarithmic factors, as well as giving various refinements that allow the algorithms to tolerate a much larger fraction of corruptions. Finally, we show on both synthetic and real data that our algorithms have state-of-the-art performance and suddenly make high-dimensional robust estimation a realistic possibility.  

Reliable Applications VI - Robustness2

3Reliable Adversarial-Examples noise Composition robustness
Presenter Papers Paper URL Our Slides
Tianlu Robustness of classifiers: from adversarial to random noise, NIPS16 PDF 1 PDF
Anant Blind Attacks on Machine Learners, 2 NIPS16 PDF PDF
  Data Noising as Smoothing in Neural Network Language Models (Ng), ICLR17 3 pdf  
  The Robustness of Estimator Composition, NIPS16 4 PDF  
  1. The Robustness of Estimator Composition, NIPS16 / We formalize notions of robustness for composite estimators via the notion of a breakdown point. A composite estimator successively applies two (or more) estimators: on data decomposed into disjoint parts, it applies the first estimator on each part, then the second estimator on the outputs of the first estimator. And so on, if the composition is of more than two estimators. Informally, the breakdown point is the minimum fraction of data points which if significantly modified will also significantly modify the output of the estimator, so it is typically desirable to have a large breakdown point. Our main result shows that, under mild conditions on the individual estimators, the breakdown point of the composite estimator is the product of the breakdown points of the individual estimators. We also demonstrate several scenarios, ranging from regression to statistical testing, where this analysis is easy to apply, useful in understanding worst case robustness, and sheds powerful insights onto the associated data analysis.  

  2. Data Noising as Smoothing in Neural Network Language Models (Ng), ICLR17/ Data noising is an effective technique for regularizing neural network models. While noising is widely adopted in application domains such as vision and speech, commonly used noising primitives have not been developed for discrete sequence-level settings such as language modeling. In this paper, we derive a connection between input noising in neural network language models and smoothing in n-gram models. Using this connection, we draw upon ideas from smoothing to develop effective noising schemes. We demonstrate performance gains when applying the proposed schemes to language modeling and machine translation. Finally, we provide empirical analysis validating the relationship between noising and smoothing.  

  3. Robustness of classifiers: from adversarial to random noise, NIPS16/ Several recent works have shown that state-of-the-art classifiers are vulnerable to worst-case (i.e., adversarial) perturbations of the datapoints. On the other hand, it has been empirically observed that these same classifiers are relatively robust to random noise. In this paper, we propose to study a semi-random noise regime that generalizes both the random and worst-case noise regimes. We propose the first quantitative analysis of the robustness of nonlinear classifiers in this general noise regime. We establish precise theoretical bounds on the robustness of classifiers in this general regime, which depend on the curvature of the classifier’s decision boundary. Our bounds confirm and quantify the empirical observations that classifiers satisfying curvature constraints are robust to random noise. Moreover, we quantify the robustness of classifiers in terms of the subspace dimension in the semi-random noise regime, and show that our bounds remarkably interpolate between the worst-case and random noise regimes. We perform experiments and show that the derived bounds provide very accurate estimates when applied to various state-of-the-art deep neural networks and datasets. This result suggests bounds on the curvature of the classifiers’ decision boundaries that we support experimentally, and more generally offers important insights onto the geometry of high dimensional classification problems.  

  4. Blind Attacks on Machine Learners, NIPS16/ The importance of studying the robustness of learners to malicious data is well established. While much work has been done establishing both robust estimators and effective data injection attacks when the attacker is omniscient, the ability of an attacker to provably harm learning while having access to little information is largely unstudied. We study the potential of a “blind attacker” to provably limit a learner’s performance by data injection attack without observing the learner’s training set or any parameter of the distribution from which it is drawn. We provide examples of simple yet effective attacks in two settings: firstly, where an “informed learner” knows the strategy chosen by the attacker, and secondly, where a “blind learner” knows only the proportion of malicious data and some family to which the malicious distribution chosen by the attacker belongs. For each attack, we analyze minimax rates of convergence and establish lower bounds on the learner’s minimax risk, exhibiting limits on a learner’s ability to learn under data injection attack even when the attacker is “blind”.  

Optimization I - Understanding DNN Optimization

4Optimization optimization Curriculum Differentiation
Presenter Papers Paper URL Our Slides
Ceyer An overview of gradient optimization algorithms, 1 PDF PDF
Shijia Osborne - Probabilistic numerics for deep learning 2 DLSS 2017 + Video PDF / PDF2
Jack Automated Curriculum Learning for Neural Networks, ICML17 3 PDF PDF
DLSS17 Johnson - Automatic Differentiation 4 slide + video  
  1. An overview of gradient optimization algorithms, by S Ruder - ‎2016 Cite as‎: ‎arXiv:1609.04747 / Gradient descent optimization algorithms, while increasingly popular, are often used as black-box optimizers, as practical explanations of their strengths and weaknesses are hard to come by. This article aims to provide the reader with intuitions with regard to the behaviour of different algorithms that will allow her to put them to use. In the course of this overview, we look at different variants of gradient descent, summarize challenges, introduce the most common optimization algorithms, review architectures in a parallel and distributed setting, and investigate additional strategies for optimizing gradient descent.  

  2. Osborne - Probabilistic numerics for deep learning / Numerical algorithms, such as methods for the numerical solution of integrals and ordinary differential equations, as well as optimization algorithms can be interpreted as estimation rules. They estimate the value of a latent, intractable quantity – the value of an integral, the solution of a differential equation, the location of an extremum – given the result of tractable computations (“observations”, such as function values of the integrand, evaluations of the differential equation, function values of the gradient of an objective). So these methods perform inference, and are accessible to the formal frameworks of probability theory. They are learning machines. Taking this observation seriously, a probabilistic numerical method is a numerical algorithm that takes in a probability distribution over its inputs, and returns a probability distribution over its output. Recent research shows that it is in fact possible to directly identify existing numerical methods, including some real classics, with specific probabilistic models. Interpreting numerical methods as learning algorithms offers various benefits. It can offer insight into the algebraic assumptions inherent in existing methods. As a joint framework for methods developed in separate communities, it allows transfer of knowledge among these areas. But the probabilistic formulation also explicitly provides a richer output than simple convergence bounds. If the probability measure returned by a probabilistic method is well-calibrated, it can be used to monitor, propagate and control the quality of computations.  

  3. Johnson - Automatic Differentiation and more: The simple essence of automatic differentiation / / Automatic differentiation (AD) in reverse mode (RAD) is a central component of deep learning and other uses of large-scale optimization. Commonly used RAD algorithms such as backpropagation, however, are complex and stateful, hindering deep understanding, improvement, and parallel execution. This paper develops a simple, generalized AD algorithm calculated from a simple, natural specification. The general algorithm is then specialized by varying the representation of derivatives. In particular, applying well-known constructions to a naive representation yields two RAD algorithms that are far simpler than previously known. In contrast to commonly used RAD implementations, the algorithms defined here involve no graphs, tapes, variables, partial derivatives, or mutation. They are inherently parallel-friendly, correct by construction, and usable directly from an existing programming language with no need for new data types or programming style, thanks to use of an AD-agnostic compiler plugin.  

  4. Automated Curriculum Learning for Neural Networks, ICML17 / lex Graves, Marc G. Bellemare, Jacob Menick, Remi Munos, Koray Kavukcuoglu/ We introduce a method for automatically selecting the path, or syllabus, that a neural network follows through a curriculum so as to maximise learning efficiency. A measure of the amount that the network learns from each data sample is provided as a reward signal to a nonstationary multi-armed bandit algorithm, which then determines a stochastic syllabus. We consider a range of signals derived from two distinct indicators of learning progress: rate of increase in prediction accuracy, and rate of increase in network complexity. Experimental results for LSTM networks on three curricula demonstrate that our approach can significantly accelerate learning, in some cases halving the time required to attain a satisfactory performance level.  

Optimization II - DNN for Optimization

4Optimization Architecture Search RL Few-Shot Optimization
Presenter Papers Paper URL Our Slides
GaoJi Neural Architecture Search with Reinforcement Learning, ICLR17 1 PDF PDF
Ceyer Learning to learn 2 DLSS17video PDF
Beilun Optimization as a Model for Few-Shot Learning, ICLR17 3 PDF + More PDF
Anant Neural Optimizer Search with Reinforcement Learning, ICML17 4 PDF PDF
  1. Neural Optimizer Search with Reinforcement Learning, ICML17 Irwan Bello, Barret Zoph, Vijay Vasudevan, Quoc V. Le / We present an approach to automate the process of discovering optimization methods, with a focus on deep learning architectures. We train a Recurrent Neural Network controller to generate a string in a domain specific language that describes a mathematical update equation based on a list of primitive functions, such as the gradient, running average of the gradient, etc. The controller is trained with Reinforcement Learning to maximize the performance of a model after a few epochs. On CIFAR-10, our method discovers several update rules that are better than many commonly used optimizers, such as Adam, RMSProp, or SGD with and without Momentum on a ConvNet model. We introduce two new optimizers, named PowerSign and AddSign, which we show transfer well and improve training on a variety of different tasks and architectures, including ImageNet classification and Google’s neural machine translation system.  

  2. Neural Architecture Search with Reinforcement Learning, ICLR17 , Barret Zoph, Quoc V. Le / Neural networks are powerful and flexible models that work well for many difficult learning tasks in image, speech and natural language understanding. Despite their success, neural networks are still hard to design. In this paper, we use a recurrent network to generate the model descriptions of neural networks and train this RNN with reinforcement learning to maximize the expected accuracy of the generated architectures on a validation set. On the CIFAR-10 dataset, our method, starting from scratch, can design a novel network architecture that rivals the best human-invented architecture in terms of test set accuracy. Our CIFAR-10 model achieves a test error rate of 3.65, which is 0.09 percent better and 1.05x faster than the previous state-of-the-art model that used a similar architectural scheme. On the Penn Treebank dataset, our model can compose a novel recurrent cell that outperforms the widely-used LSTM cell, and other state-of-the-art baselines. Our cell achieves a test set perplexity of 62.4 on the Penn Treebank, which is 3.6 perplexity better than the previous state-of-the-art model. The cell can also be transferred to the character language modeling task on PTB and achieves a state-of-the-art perplexity of 1.214.  

  3. Learning to learn / DLSS17 / Learning to learn without gradient descent by gradient descent / Yutian Chen, Matthew W Hoffman, Sergio Gómez Colmenarejo, Misha Denil, Timothy P Lillicrap, Matt Botvinick, Nando de Freitas/ We learn recurrent neural network optimizers trained on simple synthetic functions by gradient descent. We show that these learned optimizers exhibit a remarkable degree of transfer in that they can be used to efficiently optimize a broad range of derivative-free black-box functions, including Gaussian process bandits, simple control objectives, global optimization benchmarks and hyper-parameter tuning tasks. Up to the training horizon, the learned optimizers learn to tradeoff exploration and exploitation, and compare favourably with heavily engineered Bayesian optimization packages for hyper-parameter tuning.  

  4. Optimization as a Model for Few-Shot Learning, ICLR17 / achin Ravi, Hugo Larochelle/ Abstract: Though deep neural networks have shown great success in the large data domain, they generally perform poorly on few-shot learning tasks, where a model has to quickly generalize after seeing very few examples from each class. The general belief is that gradient-based optimization in high capacity models requires many iterative steps over many examples to perform well. Here, we propose an LSTM-based meta-learner model to learn the exact optimization algorithm used to train another learner neural network in the few-shot regime. The parametrization of our model allows it to learn appropriate parameter updates specifically for the scenario where a set amount of updates will be made, while also learning a general initialization of the learner network that allows for quick convergence of training. We demonstrate that this meta-learning model is competitive with deep metric-learning techniques for few-shot learning.  

Optimization III - Optimization for DNN

4Optimization Architecture-Search Hyperparameter dynamic Optimization
Presenter Papers Paper URL Our Slides
GaoJi Forward and Reverse Gradient-Based Hyperparameter Optimization, ICML17 1 PDF PDF
Chaojiang Adaptive Neural Networks for Efficient Inference, ICML17 2 PDF PDF
Bargav Practical Gauss-Newton Optimisation for Deep Learning, ICML17 3 PDF PDF
Rita How to Escape Saddle Points Efficiently, ICML17 4 PDF PDF
  Batched High-dimensional Bayesian Optimization via Structural Kernel Learning PDF  
  1. Forward and Reverse Gradient-Based Hyperparameter Optimization, ICML17/ We study two procedures (reverse-mode and forward-mode) for computing the gradient of the validation error with respect to the hyperparameters of any iterative learning algorithm such as stochastic gradient descent. These procedures mirror two methods of computing gradients for recurrent neural networks and have different trade-offs in terms of running time and space requirements. Our formulation of the reverse-mode procedure is linked to previous work by Maclaurin et al. [2015] but does not require reversible dynamics. The forward-mode procedure is suitable for real-time hyperparameter updates, which may significantly speed up hyperparameter optimization on large datasets. We present experiments on data cleaning and on learning task interactions. We also present one large-scale experiment where the use of previous gradient-based methods would be prohibitive.  

  2. Adaptive Neural Networks for Efficient Inference, ICML17 / We present an approach to adaptively utilize deep neural networks in order to reduce the evaluation time on new examples without loss of accuracy. Rather than attempting to redesign or approximate existing networks, we propose two schemes that adaptively utilize networks. We first pose an adaptive network evaluation scheme, where we learn a system to adaptively choose the components of a deep network to be evaluated for each example. By allowing examples correctly classified using early layers of the system to exit, we avoid the computational time associated with full evaluation of the network. We extend this to learn a network selection system that adaptively selects the network to be evaluated for each example. We show that computational time can be dramatically reduced by exploiting the fact that many examples can be correctly classified using relatively efficient networks and that complex, computationally costly networks are only necessary for a small fraction of examples. We pose a global objective for learning an adaptive early exit or network selection policy and solve it by reducing the policy learning problem to a layer-by-layer weighted binary classification problem. Empirically, these approaches yield dramatic reductions in computational cost, with up to a 2.8x speedup on state-of-the-art networks from the ImageNet image recognition challenge with minimal (<1%) loss of top5 accuracy.  

  3. Practical Gauss-Newton Optimisation for Deep Learning, ICML17 / We present an efficient block-diagonal approximation to the Gauss-Newton matrix for feedforward neural networks. Our resulting algorithm is competitive against state-of-the-art first order optimisation methods, with sometimes significant improvement in optimisation performance. Unlike first-order methods, for which hyperparameter tuning of the optimisation parameters is often a laborious process, our approach can provide good performance even when used with default settings. A side result of our work is that for piecewise linear transfer functions, the network objective function can have no differentiable local maxima, which may partially explain why such transfer functions facilitate effective optimisation.  

  4. How to Escape Saddle Points Efficiently, ICML17 / Chi Jin, Rong Ge, Praneeth Netrapalli, Sham M. Kakade, Michael I. Jordan/ This paper shows that a perturbed form of gradient descent converges to a second-order stationary point in a number iterations which depends only poly-logarithmically on dimension (i.e., it is almost “dimension-free”). The convergence rate of this procedure matches the well-known convergence rate of gradient descent to first-order stationary points, up to log factors. When all saddle points are non-degenerate, all second-order stationary points are local minima, and our result thus shows that perturbed gradient descent can escape saddle points almost for free. Our results can be directly applied to many machine learning applications, including deep learning. As a particular concrete example of such an application, we show that our results can be used directly to establish sharp global convergence rates for matrix factorization. Our results rely on a novel characterization of the geometry around saddle points, which may be of independent interest to the non-convex optimization community.  

Optimization IV - change DNN architecture for Optimization

4Optimization Forcing Optimization
Presenter Papers Paper URL Our Slides
Shijia Professor Forcing: A New Algorithm for Training Recurrent Networks, 1 NIPS16 PDF + Video PDF
Beilun+Arshdeep Mollifying Networks, Bengio, ICLR17 2 PDF PDF / PDF2
  1. Mollifying Networks, Bengio, ICLR17/ The optimization of deep neural networks can be more challenging than traditional convex optimization problems due to the highly non-convex nature of the loss function, e.g. it can involve pathological landscapes such as saddle-surfaces that can be difficult to escape for algorithms based on simple gradient descent. In this paper, we attack the problem of optimization of highly non-convex neural networks by starting with a smoothed – or mollified – objective function that gradually has a more non-convex energy landscape during the training. Our proposition is inspired by the recent studies in continuation methods: similar to curriculum methods, we begin learning an easier (possibly convex) objective function and let it evolve during the training, until it eventually goes back to being the original, difficult to optimize, objective function. The complexity of the mollified networks is controlled by a single hyperparameter which is annealed during the training. We show improvements on various difficult optimization tasks and establish a relationship with recent works on continuation methods for neural networks and mollifiers.  

  2. Professor Forcing: A New Algorithm for Training Recurrent Networks, NIPS16/ The Teacher Forcing algorithm trains recurrent networks by supplying observed sequence values as inputs during training and using the network’s own one-step-ahead predictions to do multi-step sampling. We introduce the Professor Forcing algorithm, which uses adversarial domain adaptation to encourage the dynamics of the recurrent network to be the same when training the network and when sampling from the network over multiple time steps. We apply Professor Forcing to language modeling, vocal synthesis on raw waveforms, handwriting generation, and image generation. Empirically we find that Professor Forcing acts as a regularizer, improving test likelihood on character level Penn Treebank and sequential MNIST. We also find that the model qualitatively improves samples, especially when sampling for a large number of time steps. This is supported by human evaluation of sample quality. Trade-offs between Professor Forcing and Scheduled Sampling are discussed. We produce T-SNEs showing that Professor Forcing successfully makes the dynamics of the network during training and sampling more similar.  

Generative II - Deep Generative Models

5Generative generative attention Composition graphical-model Autoregressive structured
Presenter Papers Paper URL Our Slides
ChaoJiang Courville - Generative Models II DLSS17Slide + video PDF
GaoJi Attend, Infer, Repeat: Fast Scene Understanding with Generative Models, NIPS16 1 PDF + talk PDF
Arshdeep Composing graphical models with neural networks for structured representations and fast inference, NIPS16 2 PDF PDF
  Johnson - Graphical Models and Deep Learning DLSSSlide + video  
  Parallel Multiscale Autoregressive Density Estimation, ICML17 3 PDF  
Beilun Conditional Image Generation with Pixel CNN Decoders, NIPS16 4 PDF PDF
Shijia Marrying Graphical Models & Deep Learning DLSS17 + Video PDF
  1. Attend, Infer, Repeat: Fast Scene Understanding with Generative Models, NIPS16 / Google DeepMind/ We present a framework for efficient inference in structured image models that explicitly reason about objects. We achieve this by performing probabilistic inference using a recurrent neural network that attends to scene elements and processes them one at a time. Crucially, the model itself learns to choose the appropriate number of inference steps. We use this scheme to learn to perform inference in partially specified 2D models (variable-sized variational auto-encoders) and fully specified 3D models (probabilistic renderers). We show that such models learn to identify multiple objects - counting, locating and classifying the elements of a scene -without any supervision, e.g., decomposing 3D images with various numbers of objects in a single forward pass of a neural network at unprecedented speed. We further show that the networks produce accurate inferences when compared to supervised counterparts, and that their structure leads to improved generalization.  

  2. Composing graphical models with neural networks for structured representations and fast inference, NIPS16 / We propose a general modeling and inference framework that combines the complementary strengths of probabilistic graphical models and deep learning methods. Our model family composes latent graphical models with neural network observation likelihoods. For inference, we use recognition networks to produce local evidence potentials, then combine them with the model distribution using efficient message-passing algorithms. All components are trained simultaneously with a single stochastic variational inference objective. We illustrate this framework by automatically segmenting and categorizing mouse behavior from raw depth video, and demonstrate several other example models.  

  3. Parallel Multiscale Autoregressive Density Estimation, ICML17 / , Nando de Freitas/ PixelCNN achieves state-of-the-art results in density estimation for natural images. Although training is fast, inference is costly, requiring one network evaluation per pixel; O(N) for N pixels. This can be sped up by caching activations, but still involves generating each pixel sequentially. In this work, we propose a parallelized PixelCNN that allows more efficient inference by modeling certain pixel groups as conditionally independent. Our new PixelCNN model achieves competitive density estimation and orders of magnitude speedup - O(log N) sampling instead of O(N) - enabling the practical generation of 512x512 images. We evaluate the model on class-conditional image generation, text-to-image synthesis, and action-conditional video generation, showing that our model achieves the best results among non-pixel-autoregressive density models that allow efficient sampling.  

  4. Conditional Image Generation with Pixel CNN Decoders, NIPS16 / Google DeepMind/ This work explores conditional image generation with a new image density model based on the PixelCNN architecture. The model can be conditioned on any vector, including descriptive labels or tags, or latent embeddings created by other networks. When conditioned on class labels from the ImageNet database, the model is able to generate diverse, realistic scenes representing distinct animals, objects, landscapes and structures. When conditioned on an embedding produced by a convolutional network given a single image of an unseen face, it generates a variety of new portraits of the same person with different facial expressions, poses and lighting conditions. We also show that conditional PixelCNN can serve as a powerful decoder in an image autoencoder. Additionally, the gated convolutional layers in the proposed model improve the log-likelihood of PixelCNN to match the state-of-the-art performance of PixelRNN on ImageNet, with greatly reduced computational cost.  

Generative III - GAN training

5Generative generative generalization Denoising Model-Criticism
Presenter Papers Paper URL Our Slides
Arshdeep Generalization and Equilibrium in Generative Adversarial Nets (ICML17) 1 PDF + video PDF
Arshdeep Mode Regularized Generative Adversarial Networks (ICLR17) 2 PDF PDF
Bargav Improving Generative Adversarial Networks with Denoising Feature Matching, ICLR17 3 PDF PDF
Anant Generative Models and Model Criticism via Optimized Maximum Mean Discrepancy, ICLR17 4 PDF + code PDF
  1. Generalization and Equilibrium in Generative Adversarial Nets (ICML17)/ We show that training of generative adversarial network (GAN) may not have good generalization properties; e.g., training may appear successful but the trained distribution may be far from target distribution in standard metrics. However, generalization does occur for a weaker metric called neural net distance. It is also shown that an approximate pure equilibrium exists in the discriminator/generator game for a special class of generators with natural training objectives when generator capacity and training set sizes are moderate. This existence of equilibrium inspires MIX+GAN protocol, which can be combined with any existing GAN training, and empirically shown to improve some of them.  

  2. Mode Regularized Generative Adversarial Networks (ICLR17)/ Although Generative Adversarial Networks achieve state-of-the-art results on a variety of generative tasks, they are regarded as highly unstable and prone to miss modes. We argue that these bad behaviors of GANs are due to the very particular functional shape of the trained discriminators in high dimensional spaces, which can easily make training stuck or push probability mass in the wrong direction, towards that of higher concentration than that of the data generating distribution. We introduce several ways of regularizing the objective, which can dramatically stabilize the training of GAN models. We also show that our regularizers can help the fair distribution of probability mass across the modes of the data generating distribution, during the early phases of training and thus providing a unified solution to the missing modes problem.  

  3. Improving Generative Adversarial Networks with Denoising Feature Matching, ICLR17 / David Warde-Farley, Yoshua Bengio/ Abstract: We propose an augmented training procedure for generative adversarial networks designed to address shortcomings of the original by directing the generator towards probable configurations of abstract discriminator features. We estimate and track the distribution of these features, as computed from data, with a denoising auto-encoder, and use it to propose high-level targets for the generator. We combine this new loss with the original and evaluate the hybrid criterion on the task of unsupervised image synthesis from datasets comprising a diverse set of visual categories, noting a qualitative and quantitative improvement in the objectness’’ of the resulting samples.  

  4. Generative Models and Model Criticism via Optimized Maximum Mean Discrepancy, ICLR17 / Dougal J. Sutherland, Hsiao-Yu Tung, Heiko Strathmann, Soumyajit De, Aaditya Ramdas, Alex Smola, Arthur Gretton / We propose a method to optimize the representation and distinguishability of samples from two probability distributions, by maximizing the estimated power of a statistical test based on the maximum mean discrepancy (MMD). This optimized MMD is applied to the setting of unsupervised learning by generative adversarial networks (GAN), in which a model attempts to generate realistic samples, and a discriminator attempts to tell these apart from data samples. In this context, the MMD may be used in two roles: first, as a discriminator, either directly on the samples, or on features of the samples. Second, the MMD can be used to evaluate the performance of a generative model, by testing the model’s samples against a reference data set. In the latter role, the optimized MMD is particularly helpful, as it gives an interpretable indication of how the model and data distributions differ, even in cases where individual model samples are not easily distinguished either by eye or by classifier.  

RL II - Basic tutorial RLSS17

6Reinforcement RL Multi-Task
Presenter Papers Paper URL Our Slides
Jack Hasselt - Deep Reinforcement Learning RLSS17.pdf + video PDF
Tianlu Roux - RL in the Industry RLSS17.pdf + video PDF / PDF-Bandit
Xueying Singh - Steps Towards Continual Learning pdf + video PDF
GaoJi Distral: Robust Multitask Reinforcement Learning 1 PDF PDF
  1. Distral: Robust Multitask Reinforcement Learning / 2017 Yee Whye Teh, Victor Bapst, Wojciech Marian Czarnecki, John Quan, James Kirkpatrick, Raia Hadsell, Nicolas Heess, Razvan Pascanu/ Most deep reinforcement learning algorithms are data inefficient in complex and rich environments, limiting their applicability to many scenarios. One direction for improving data efficiency is multitask learning with shared neural network parameters, where efficiency may be improved through transfer across related tasks. In practice, however, this is not usually observed, because gradients from different tasks can interfere negatively, making learning unstable and sometimes even less data efficient. Another issue is the different reward schemes between tasks, which can easily lead to one task dominating the learning of a shared model. We propose a new approach for joint training of multiple tasks, which we refer to as Distral (Distill & transfer learning). Instead of sharing parameters between the different workers, we propose to share a “distilled” policy that captures common behaviour across tasks. Each worker is trained to solve its own task while constrained to stay close to the shared policy, while the shared policy is trained by distillation to be the centroid of all task policies. Both aspects of the learning process are derived by optimizing a joint objective function. We show that our approach supports efficient transfer on complex 3D environments, outperforming several related methods. Moreover, the proposed learning process is more robust and more stable—attributes that are critical in deep reinforcement learning.  

RL III - Basic tutorial RLSS17 (2)

6Reinforcement alphaGO Planning Temporal-Difference
Presenter Papers Paper URL Our Slides
Anant The Predictron: End-to-End Learning and Planning, ICLR17 1 PDF PDF
ChaoJiang Szepesvari - Theory of RL 2 RLSS.pdf + Video PDF
GaoJi Mastering the game of Go without human knowledge / Nature 2017 3 PDF PDF
  Thomas - Safe Reinforcement Learning RLSS17.pdf + video  
  Sutton - Temporal-Difference Learning RLSS17.pdf + Video  
  1. Temporal-Difference Learning basics / / Temporal difference (TD) learning refers to a class of model-free reinforcement learning methods which learn by bootstrapping from the current estimate of the value function. These methods sample from the environment, like Monte Carlo methods, and perform updates based on current estimates, like dynamic programming methods. While Monte Carlo methods only adjust their estimates once the final outcome is known, TD methods adjust predictions to match later, more accurate, predictions about the future before the final outcome is known.  

  2. The Predictron: End-to-End Learning and Planning, ICLR17 / David Silver/ One of the key challenges of artificial intelligence is to learn models that are effective in the context of planning. In this document we introduce the predictron architecture. The predictron consists of a fully abstract model, represented by a Markov reward process, that can be rolled forward multiple “imagined” planning steps. Each forward pass of the predictron accumulates internal rewards and values over multiple planning depths. The predictron is trained end-to-end so as to make these accumulated values accurately approximate the true value function. We applied the predictron to procedurally generated random mazes and a simulator for the game of pool. The predictron yielded significantly more accurate predictions than conventional deep neural network architectures.  

  3. Mastering the game of Go without human knowledge / Nature 2017 Google DeepMind / A long-standing goal of artificial intelligence is an algorithm that learns, tabula rasa, superhuman proficiency in challenging domains. Recently, AlphaGo became the first program to defeat a world champion in the game of Go. The tree search in AlphaGo evaluated positions and selected moves using deep neural networks. These neural networks were trained by supervised learning from human expert moves, and by reinforcement learning from self-play. Here we introduce an algorithm based solely on reinforcement learning, without human data, guidance or domain knowledge beyond game rules. AlphaGo becomes its own teacher: a neural network is trained to predict AlphaGo’s own move selections and also the winner of AlphaGo’s games. This neural network improves the strength of the tree search, resulting in higher quality move selection and stronger self-play in the next iteration. Starting tabula rasa, our new program AlphaGo Zero achieved superhuman performance, winning 100–0 against the previously published, champion-defeating AlphaGo.  

RL IV - RL with varying structures

6Reinforcement Auxiliary Sampling Value-Networks structured Imitation-Learning Hierarchical
Presenter Papers Paper URL Our Slides
Ceyer Reinforcement Learning with Unsupervised Auxiliary Tasks, ICLR17 1 PDF PDF
Beilun Why is Posterior Sampling Better than Optimism for Reinforcement Learning? Ian Osband, Benjamin Van Roy 2 PDF PDF
Ji Deeply AggreVaTeD: Differentiable Imitation Learning for Sequential Prediction, ICML17 3 PDF PDF
Xueying End-to-End Differentiable Adversarial Imitation Learning, ICML17 4 PDF PDF
  Deep Value Networks Learn to Evaluate and Iteratively Refine Structured Outputs, ICML17 PDF  
  FeUdal Networks for Hierarchical Reinforcement Learning, ICML17 5 PDF  
  1. Reinforcement Learning with Unsupervised Auxiliary Tasks, ICLR17 / Max Jaderberg, Volodymyr Mnih, Wojciech Marian Czarnecki, Tom Schaul, Joel Z Leibo, David Silver, Koray Kavukcuoglu/ Deep reinforcement learning agents have achieved state-of-the-art results by directly maximising cumulative reward. However, environments contain a much wider variety of possible training signals. In this paper, we introduce an agent that also maximises many other pseudo-reward functions simultaneously by reinforcement learning. All of these tasks share a common representation that, like unsupervised learning, continues to develop in the absence of extrinsic rewards. We also introduce a novel mechanism for focusing this representation upon extrinsic rewards, so that learning can rapidly adapt to the most relevant aspects of the actual task. Our agent significantly outperforms the previous state-of-the-art on Atari, averaging 880\% expert human performance, and a challenging suite of first-person, three-dimensional \emph{Labyrinth} tasks leading to a mean speedup in learning of 10× and averaging 87\% expert human performance on Labyrinth.  

  2. Deeply AggreVaTeD: Differentiable Imitation Learning for Sequential Prediction, ICML17/ Researchers have demonstrated state-of-the-art performance in sequential decision making problems (e.g., robotics control, sequential prediction) with deep neural network models. One often has access to near-optimal oracles that achieve good performance on the task during training. We demonstrate that AggreVaTeD — a policy gradient extension of the Imitation Learning (IL) approach of (Ross & Bagnell, 2014) — can leverage such an oracle to achieve faster and better solutions with less training data than a less-informed Reinforcement Learning (RL) technique. Using both feedforward and recurrent neural network predictors, we present stochastic gradient procedures on a sequential prediction task, dependency-parsing from raw image data, as well as on various high dimensional robotics control problems. We also provide a comprehensive theoretical study of IL that demonstrates we can expect up to exponentially lower sample complexity for learning with AggreVaTeD than with RL algorithms, which backs our empirical findings. Our results and theory indicate that the proposed approach can achieve superior performance with respect to the oracle when the demonstrator is sub-optimal.  

  3. End-to-End Differentiable Adversarial Imitation Learning, ICML17/ Generative Adversarial Networks (GANs) have been successfully applied to the problem of policy imitation in a model-free setup. However, the computation graph of GANs, that include a stochastic policy as the generative model, is no longer differentiable end-to-end, which requires the use of high-variance gradient estimation. In this paper, we introduce the Model-based Generative Adversarial Imitation Learning (MGAIL) algorithm. We show how to use a forward model to make the computation fully differentiable, which enables training policies using the exact gradient of the discriminator. The resulting algorithm trains competent policies using relatively fewer expert samples and interactions with the environment. We test it on both discrete and continuous action domains and report results that surpass the state-of-the-art.  

  4. Deep Value Networks Learn to Evaluate and Iteratively Refine Structured Outputs, ICML17/ We approach structured output prediction by optimizing a deep value network (DVN) to precisely estimate the task loss on different output configurations for a given input. Once the model is trained, we perform inference by gradient descent on the continuous relaxations of the output variables to find outputs with promising scores from the value network. When applied to image segmentation, the value network takes an image and a segmentation mask as inputs and predicts a scalar estimating the intersection over union between the input and ground truth masks. For multi-label classification, the DVN’s objective is to correctly predict the F1 score for any potential label configuration. The DVN framework achieves the state-of-the-art results on multi-label prediction and image segmentation benchmarks.  

  5. FeUdal Networks for Hierarchical Reinforcement Learning, ICML17 / Alexander Sasha Vezhnevets, Simon Osindero, Tom Schaul, Nicolas Heess, Max Jaderberg, David Silver, Koray Kavukcuoglu/ We introduce FeUdal Networks (FuNs): a novel architecture for hierarchical reinforcement learning. Our approach is inspired by the feudal reinforcement learning proposal of Dayan and Hinton, and gains power and efficacy by decoupling end-to-end learning across multiple levels – allowing it to utilise different resolutions of time. Our framework employs a Manager module and a Worker module. The Manager operates at a lower temporal resolution and sets abstract goals which are conveyed to and enacted by the Worker. The Worker generates primitive actions at every tick of the environment. The decoupled structure of FuN conveys several benefits – in addition to facilitating very long timescale credit assignment it also encourages the emergence of sub-policies associated with different goals set by the Manager. These properties allow FuN to dramatically outperform a strong baseline agent on tasks that involve long-term credit assignment or memorisation. We demonstrate the performance of our proposed system on a range of tasks from the ATARI suite and also from a 3D DeepMind Lab environment.