text
stringlengths 23
5.09k
| label
class label 3
classes |
|---|---|
Title: Prototypical Examples in Deep Learning: Metrics, Characteristics, and Utility. Abstract: Machine learning (ML) research has investigated prototypes: examples that are representative of the behavior to be learned. We systematically evaluate five methods for identifying prototypes, both ones previously introduced as well as new ones we propose, finding all of them to provide meaningful but different interpretations. Through a human study, we confirm that all five metrics are well matched to human intuition. Examining cases where the metrics disagree offers an informative perspective on the properties of data and algorithms used in learning, with implications for data-corpus construction, efficiency, adversarial robustness, interpretability, and other ML aspects. In particular, we confirm that the "train on hard" curriculum approach can improve accuracy on many datasets and tasks, but that it is strictly worse when there are many mislabeled or ambiguous examples. | 0reject
|
Title: Composing Features: Compositional Model Augmentation for Steerability of Music Transformers. Abstract: Music is a combinatorial art. Given a starting sequence, many continuations are possible, yet often only one is written down. With generative models, we can explore many. However, finding a continuation with specific combinations of features (such as rising pitches, with block chords played in syncopated rhythm) can take many trials.
To tackle the combinatorial nature of composing features, we propose a compositional approach to steering music transformers, building on lightweight fine-tuning methods such as prefix tuning and bias tuning. We introduce a novel contrastive loss function that enables us to steer compositional models over logical features using supervised learning. We examine the difficulty in steering based on whether features musically follow a prime or not, using existing music as a proxy. We show that with a relatively small number of extra parameters, our method allows bias tuning to perform successful fine-tuning in both the single-feature and compositional setting. | 0reject
|
Title: On the Implicit Biases of Architecture & Gradient Descent. Abstract: Do neural networks generalise because of bias in the functions returned by gradient descent, or bias already present in the network architecture? $\textit{¿Por qué no los dos?}$ This paper finds that while typical networks that fit the training data already generalise fairly well, gradient descent can further improve generalisation by selecting networks with a large margin. This conclusion is based on a careful study of the behaviour of infinite width networks trained by Bayesian inference and finite width networks trained by gradient descent. To measure the implicit bias of architecture, new technical tools are developed to both $\textit{analytically bound}$ and $\textit{consistently estimate}$ the average test error of the neural network--Gaussian process (NNGP) posterior. This error is found to be already better than chance, corroborating the findings of Valle-Pérez et al. (2019) and underscoring the importance of architecture. Going beyond this result, this paper finds that test performance can be substantially improved by selecting a function with much larger margin than is typical under the NNGP posterior. This highlights a curious fact: $\textit{minimum a posteriori}$ functions can generalise best, and gradient descent can select for those functions. In summary, new technical tools suggest a nuanced portrait of generalisation involving both the implicit biases of architecture and gradient descent. | 0reject
|
Title: Palette: Image-to-Image Diffusion Models. Abstract: We introduce Palette, a simple and general framework for image-to-image translation using conditional diffusion models. On four challenging image-to-image translation tasks (colorization, inpainting, uncropping, and JPEG decompression), Palette outperforms strong GAN and regression baselines, and establishes a new state-of-the-art result. This is accomplished without task-specific hyper-parameter tuning, architecture customization, or any auxiliary loss demonstrating a desirable degree of generality and flexibility. We uncover the impact of using L2 vs. L1 loss in the denoising diffusion objective on sample diversity, and demonstrate the importance of self-attention through empirical architecture studies. Importantly, we advocate a unified evaluation protocol based on ImageNet, and report several sample quality scores including FID, Inception Score, Classification Accuracy of a pre-trained ResNet-50, and Perceptual Distance against reference images for various baselines. We expect this standardized evaluation protocol to play a critical role in advancing image-to-image translation research. Finally, we show that a single generalist Palette model trained on 3 tasks (colorization, inpainting, JPEG decompression) performs as well or better than task-specific specialist counterparts. | 2withdrawn
|
Title: Can Deep Reinforcement Learning solve Erdos-Selfridge-Spencer Games?. Abstract: Deep reinforcement learning has achieved many recent successes, but our understanding of its strengths and limitations is hampered by the lack of rich environments in which we can fully characterize optimal behavior, and correspondingly diagnose individual actions against such a characterization.
Here we consider a family of combinatorial games, arising from work of Erdos, Selfridge, and Spencer, and we propose their use as environments for evaluating and comparing different approaches to reinforcement learning. These games have a number of appealing features: they are challenging for current learning approaches, but they form (i) a low-dimensional, simply parametrized environment where (ii) there is a linear closed form solution for optimal behavior from any state, and (iii) the difficulty of the game can be tuned by changing environment parameters in an interpretable way. We use these Erdos-Selfridge-Spencer games not only to compare different algorithms, but also to compare approaches based on supervised and reinforcement learning, to analyze the power of multi-agent approaches in improving performance, and to evaluate generalization to environments outside the training set. | 0reject
|
Title: Multi-Task Learning by a Top-Down Control Network. Abstract: As the range of tasks performed by a general vision system expands, executing multiple tasks accurately and efficiently in a single network has become an important and still open problem. Recent computer vision approaches address this problem by branching networks, or by a channel-wise modulation of the network feature-maps with task specific vectors. We present a novel architecture that uses a dedicated top-down control network to modify the activation of all the units in the main recognition network in a manner that depends on the selected task, image content, and spatial location. We show the effectiveness of our scheme by achieving significantly better results than alternative state-of-the-art approaches on four datasets. We further demonstrate our advantages in terms of task selectivity, scaling the number of tasks and interpretability.
Code is supplied in the supplementary materials and will be publicly available. | 0reject
|
Title: Representational correlates of hierarchical phrase structure in deep language models. Abstract: While contextual representations from Transformer-based architectures have set a new standard for many NLP tasks, there is not yet a complete accounting of their inner workings. In particular, it is not entirely clear what aspects of sentence-level syntax are captured by these representations, nor how (if at all) they are built along the stacked layers of the network. In this paper, we aim to address such questions with a general class of input perturbation-based analyses of representations from Transformer networks pretrained on self-supervised objectives. Importing from computational and cognitive neuroscience the notion of representational invariance, we perform a series of probes designed to test the sensitivity of Transformer representations to several kinds of structure in sentences. Each probe involves swapping words in a sentence and comparing the representations from perturbed sentences against the original. We experiment with three different perturbations: (1) random permutations of n-grams of varying width, to test the scale at which a representation is sensitive to word position; (2) swapping of two spans which do or do not form a syntactic phrase, to test sensitivity to global phrase structure; and (3) swapping of two adjacent words which do or do not break apart a syntactic phrase, to test sensitivity to local phrase structure. We also connect our probe results to the Transformer architecture by relating the attention mechanism to syntactic distance between two words. Results from the three probes collectively suggest that Transformers build sensitivity to larger parts of the sentence along their layers, and that hierarchical phrase structure plays a role in this process. In particular, sensitivity to local phrase structure increases along deeper layers. Based on our analysis of attention, we show that this is at least partly explained by generally larger attention weights between syntactically distant words. | 0reject
|
Title: Unsupervised-Learning of time-varying features. Abstract: We present an architecture based on the conditional Variational Autoencoder to learn a representation
of transformations in time-sequence data. The model is constructed in a way that allows to identify sub-spaces of features indicating changes between frames without learning features that are constant within a time-sequence. Therefore, the approach disentangles content from transformations. Different model-architectures are applied to affine image-transformations on MNIST as well as a car-racing video-game task.
Results show that the model discovers relevant parameterizations, however, model architecture has a major impact on the feature-space. It turns out, that there is an advantage of only learning features describing change of state between images, over learning the states of the images at each frame. In this case, we do not only achieve higher accuracy but also more interpretable linear features. Our results also uncover the need for model architectures that combine global transformations with convolutional architectures. | 0reject
|
Title: On the Stability of Fine-tuning BERT: Misconceptions, Explanations, and Strong Baselines. Abstract: Fine-tuning pre-trained transformer-based language models such as BERT has become a common practice dominating leaderboards across various NLP benchmarks. Despite the strong empirical performance of fine-tuned models, fine-tuning is an unstable process: training the same model with multiple random seeds can result in a large variance of the task performance. Previous literature (Devlin et al., 2019; Lee et al., 2020; Dodge et al., 2020) identified two potential reasons for the observed instability: catastrophic forgetting and small size of the fine-tuning datasets. In this paper, we show that both hypotheses fail to explain the fine-tuning instability. We analyze BERT, RoBERTa, and ALBERT, fine-tuned on commonly used datasets from the GLUE benchmark, and show that the observed instability is caused by optimization difficulties that lead to vanishing gradients. Additionally, we show that the remaining variance of the downstream task performance can be attributed to differences in generalization where fine-tuned models with the same training loss exhibit noticeably different test performance. Based on our analysis, we present a simple but strong baseline that makes fine-tuning BERT-based models significantly more stable than the previously proposed approaches. Code to reproduce our results is available online: https://github.com/uds-lsv/bert-stable-fine-tuning. | 1accept
|
Title: Understanding and Leveraging Overparameterization in Recursive Value Estimation. Abstract: The theory of function approximation in reinforcement learning (RL) typically considers low capacity representations that incur a tradeoff between approximation error, stability and generalization. Current deep architectures, however, operate in an overparameterized regime where approximation error is not necessarily a bottleneck. To better understand the utility of deep models in RL we present an analysis of recursive value estimation using \emph{overparameterized} linear representations that provides useful, transferable findings. First, we show that classical updates such as temporal difference (TD) learning or fitted-value-iteration (FVI) converge to \emph{different} fixed points than residual minimization (RM) in the overparameterized linear case. We then develop a unified interpretation of overparameterized linear value estimation as minimizing the Euclidean norm of the weights subject to alternative constraints. A practical consequence is that RM can be modified by a simple alteration of the backup targets to obtain the same fixed points as FVI and TD (when they converge), while universally ensuring stability. Further, we provide an analysis of the generalization error of these methods, demonstrating per iterate bounds on the value prediction error of FVI, and fixed point bounds for TD and RM.
Given this understanding, we then develop new algorithmic tools for improving recursive value estimation with deep models.
In particular, we extract two regularizers that penalize out-of-span top-layer weights and co-linearity in top-layer features respectively. Empirically we find that these regularizers dramatically improve the stability of TD and FVI, while allowing RM to match and even sometimes surpass their generalization performance with assured stability. | 1accept
|
Title: Unsupervised Domain Adaptation Via Pseudo-labels And Objectness Constraints. Abstract: Pseudo label self-training has emerged as a dominant approach to unsupervised domain adaptation (UDA) for semantic segmentation. Despite recent advances, this approach is susceptible to erroneous pseudo labels arising from confirmation bias that ultimately leads to sub-optimal segmentation. To mitigate the effect of noisy pseudo-labels, we propose regularising conventional self-training objectives with constraints that are derived from structure-preserving modalities, such as depth. Towards this end, we introduce a contrastive image-level objectness constraint that pulls the pixel representations of the same object instance closer while pushing those from different object categories apart. To identify pixels within an object, we subscribe to a notion of objectness derived from depth maps, that are robust to photometric variations, as well as superpixels, that are obtained via unsupervised clustering over the raw image space. Crucially, the objectness constraint is agnostic to the ground-truth semantic segmentation labels and, therefore, remains appropriate for unsupervised adaptation settings. In this paper, we show that our approach of leveraging multi-modal constraint improves top performing self-training methods in various UDA benchmarks for semantic segmentation. We make our code and data-splits available in the supplementary material. | 2withdrawn
|
Title: You Only Sample (Almost) Once: Linear Cost Self-Attention Via Bernoulli Sampling. Abstract: Transformer-based models have come to dominate the landscape in a wide range of natural language processing (NLP) applications. The heart of the transformer model is the self-attention mechanism, which captures the interactions of token pairs in the input sequences and consequently, depends quadratically on the input sequence length. It is known that training such models on longer sequences is quite expensive, and often, prohibitively so. We show that a Bernoulli sampling attention mechanism based on Locality Sensitive Hashing (LSH), decreases the quadratic complexity to linear. We bypass the quadratic cost by considering self-attention as a sum of individual tokens associated with Bernoulli random variables that can, in principle, be sampled at once by a single hash (although in practice, this number may be a small constant). This leads to an efficient sampling scheme to estimate self-attention which relies on specific modifications of LSH (based on feasibility of deployment on GPU architectures). We evaluate our proposed algorithm on the GLUE benchmark with standard 512 sequence length and our method achieves comparable or even slightly better performance than a standard pretrained Transformer. To evaluate whether our method can indeed handle longer sequences, we conduct experiments on long sequence (4096) language model pretraining and achieve consistent results as standard self-attention, while observing sizable inference speed-ups and memory savings. | 0reject
|
Title: SALSA-TEXT : SELF ATTENTIVE LATENT SPACE BASED ADVERSARIAL TEXT GENERATION. Abstract: Inspired by the success of self attention mechanism and Transformer architecture
in sequence transduction and image generation applications, we propose novel self
attention-based architectures to improve the performance of adversarial latent code-
based schemes in text generation. Adversarial latent code-based text generation
has recently gained a lot of attention due to their promising results. In this paper,
we take a step to fortify the architectures used in these setups, specifically AAE
and ARAE. We benchmark two latent code-based methods (AAE and ARAE)
designed based on adversarial setups. In our experiments, the Google sentence
compression dataset is utilized to compare our method with these methods using
various objective and subjective measures. The experiments demonstrate the
proposed (self) attention-based models outperform the state-of-the-art in adversarial
code-based text generation. | 2withdrawn
|
Title: TopicGAN: Unsupervised Text Generation from Explainable Latent Topics. Abstract: Learning discrete representations of data and then generating data from the discovered representations have been increasingly studied because the obtained discrete representations can benefit unsupervised learning. However, the performance of learning discrete representations of textual data with deep generative models has not been widely explored. In addition, although generative adversarial networks(GAN) have shown impressing results in many areas such as image generation, for text generation, it is notorious for extremely difficult to train. In this work, we propose TopicGAN, a two-step text generative model, which is able to solve those two important problems simultaneously. In the first step, it discovers the latent topics and produced bag-of-words according to the latent topics. In the second step, it generates text from the produced bag-of-words. In our experiments, we show our model can discover meaningful discrete latent topics of texts in an unsupervised fashion and generate high quality natural language from the discovered latent topics. | 0reject
|
Title: Beyond Trivial Counterfactual Generations with Diverse Valuable Explanations. Abstract: Explainability of machine learning models has gained considerable attention within our research community given the importance of deploying more reliable machine-learning systems. Explanability can also be helpful for model debugging. In computer vision applications, most methods explain models by displaying the regions in the input image that they focus on for their prediction, but it is difficult to improve models based on these explanations since they do not indicate why the model fail. Counterfactual methods, on the other hand, indicate how to perturb the input to change the model prediction, providing details about the model's decision-making. Unfortunately, current counterfactual methods make ambiguous interpretations as they combine multiple biases of the model and the data in a single counterfactual interpretation of the model's decision. Moreover, these methods tend to generate trivial counterfactuals about the model's decision, as they often suggest to exaggerate or remove the presence of the attribute being classified. Trivial counterfactuals are usually not valuable, since the information they provide is often already known to the system's designer. In this work, we propose a counterfactual method that learns a perturbation in a disentangled latent space that is constrained using a diversity-enforcing loss to uncover multiple valuable explanations about the model's prediction. Further, we introduce a mechanism to prevent the model from producing trivial explanations. Experiments on CelebA and Synbols demonstrate that our model improves the success rate of producing high-quality valuable explanations when compared to previous state-of-the-art methods. We will make the code public. | 0reject
|
Title: Inverse Constrained Reinforcement Learning. Abstract: Standard reinforcement learning (RL) algorithms train agents to maximize given reward functions. However, many real-world applications of RL require agents to also satisfy certain constraints which may, for example, be motivated by safety concerns. Constrained RL algorithms approach this problem by training agents to maximize given reward functions while respecting \textit{explicitly} defined constraints. However, in many cases, manually designing accurate constraints is a challenging task. In this work, given a reward function and a set of demonstrations from an expert that maximizes this reward function while respecting \textit{unknown} constraints, we propose a framework to learn the most likely constraints that the expert respects. We then train agents to maximize the given reward function subject to the learned constraints. Previous works in this regard have either mainly been restricted to tabular settings or specific types of constraints or assume knowledge of transition dynamics of the environment. In contrast, we empirically show that our framework is able to learn arbitrary \textit{Markovian} constraints in high-dimensions in a model-free setting. | 0reject
|
Title: Feature Interaction Interpretability: A Case for Explaining Ad-Recommendation Systems via Neural Interaction Detection. Abstract: Recommendation is a prevalent application of machine learning that affects many users; therefore, it is important for recommender models to be accurate and interpretable. In this work, we propose a method to both interpret and augment the predictions of black-box recommender systems. In particular, we propose to interpret feature interactions from a source recommender model and explicitly encode these interactions in a target recommender model, where both source and target models are black-boxes. By not assuming the structure of the recommender system, our approach can be used in general settings. In our experiments, we focus on a prominent use of machine learning recommendation: ad-click prediction. We found that our interaction interpretations are both informative and predictive, e.g., significantly outperforming existing recommender models. What's more, the same approach to interpret interactions can provide new insights into domains even beyond recommendation, such as text and image classification. | 1accept
|
Title: Generalizing Reinforcement Learning to Unseen Actions. Abstract: A fundamental trait of intelligence is the ability to achieve goals in the face of novel circumstances. In this work, we address one such setting which requires solving a task with a novel set of actions. Empowering machines with this ability requires generalization in the way an agent perceives its available actions along with the way it uses these actions to solve tasks. Hence, we propose a framework to enable generalization over both these aspects: understanding an action’s functionality, and using actions to solve tasks through reinforcement learning. Specifically, an agent interprets an action’s behavior using unsupervised representation learning over a collection of data samples reflecting the diverse properties of that action. We employ a reinforcement learning architecture which works over these action representations, and propose regularization metrics essential for enabling generalization in a policy. We illustrate the generalizability of the representation learning method and policy, to enable zero-shot generalization to previously unseen actions on challenging sequential decision-making environments. Our results and videos can be found at sites.google.com/view/action-generalization/ | 0reject
|
Title: Imitation Learning via Off-Policy Distribution Matching. Abstract: When performing imitation learning from expert demonstrations, distribution matching is a popular approach, in which one alternates between estimating distribution ratios and then using these ratios as rewards in a standard reinforcement learning (RL) algorithm. Traditionally, estimation of the distribution ratio requires on-policy data, which has caused previous work to either be exorbitantly data- inefficient or alter the original objective in a manner that can drastically change its optimum. In this work, we show how the original distribution ratio estimation objective may be transformed in a principled manner to yield a completely off-policy objective. In addition to the data-efficiency that this provides, we are able to show that this objective also renders the use of a separate RL optimization unnecessary. Rather, an imitation policy may be learned directly from this objective without the use of explicit rewards. We call the resulting algorithm ValueDICE and evaluate it on a suite of popular imitation learning benchmarks, finding that it can achieve state-of-the-art sample efficiency and performance. | 1accept
|
Title: Improving Discriminative Visual Representation Learning via Automatic Mixup. Abstract: Mixup, a convex interpolation technique for data augmentation, has achieved great success in deep neural networks. However, the community usually confines it to supervised scenarios or applies it as a predefined augmentation strategy in various fields, grossly underestimating its capacity for modeling relationships between two classes or instances. In this paper, we decompose mixup into two sub-tasks of mixup generation and classification and formulate it for discriminative representations as class- and instance-level mixup. We first analyze and summarize the properties of instance-level mixup as local smoothness and global discrimination. Then, we improve mixup generation with these properties from two aspects: we enhance modeling non-linear mixup relationships between two samples and discuss learning objectives for mixup generation. Eventually, we propose a general mixup training method called AMix to improve discriminative representations on various scenarios. Extensive experiments on supervised and self-supervised scenarios show that AMix consistently outperforms leading methods by a large margin. | 2withdrawn
|
Title: Towards Robust and Efficient Contrastive Textual Representation Learning. Abstract: There has been growing interest in representation learning for text data, based on theoretical arguments and empirical evidence. One important direction involves leveraging contrastive learning to improve learned representations. We propose an application of contrastive learning for intermediate textual feature pairs, to explicitly encourage the model to learn more distinguishable representations. To overcome the learner's degeneracy due to vanishing contrasting signals, we impose Wasserstein constraints on the critic via spectral regularization.
Finally, to moderate such an objective from overly regularized training and to enhance learning efficiency, with theoretical justification, we further leverage an active negative-sample-selection procedure to only use high-quality contrast examples. We evaluate the proposed method over a wide range of natural language processing applications, from the perspectives of both supervised and unsupervised learning. Empirical results show consistent improvement over baselines. | 0reject
|
Title: General Adversarial Defense via Pixel Level and Feature Level Distribution Alignment. Abstract: Deep neural networks (DNNs) have achieved amazing success on a wide range of high-level computer vision tasks. However, it is proved that DNNs are vulnerable to adversarial samples. The threat of adversarial samples comes from the large distribution gap between adversarial samples and clean samples in the feature spaces of the target DNNs. To this, we utilize deep generative networks with a novel training scheme to eliminate the distribution gap. Our training strategy introduces constraints in both pixel level as well as feature level, and the trained network can effectively align the distribution of adversarial samples with clean samples for target DNNs through translating their pixel values. Specifically, compared with previous methods, we propose a more efficient pixel-level training constraint to weaken the hardness of aligning adversarial samples to clean samples, which can thus obviously enhance the robustness on adversarial samples. Besides, a class-aware feature-level constraint is formulated for integrated distribution alignment. Our approach is general and suitable for multiple tasks like image classification, semantic segmentation and object detection. We conduct extensive experiments on these three tasks and different datasets, on which the superiority of our strategy over existing methods demonstrates its effectiveness and generality. | 2withdrawn
|
Title: Learning temporal evolution of probability distribution with Recurrent Neural Network. Abstract: We propose to tackle a time series regression problem by computing temporal evolution of a probability density function to provide a probabilistic forecast. A Recurrent Neural Network (RNN) based model is employed to learn a nonlinear operator for temporal evolution of a probability density function. We use a softmax layer for a numerical discretization of a smooth probability density functions, which transforms a function approximation problem to a classification task. Explicit and implicit regularization strategies are introduced to impose a smoothness condition on the estimated probability distribution. A Monte Carlo procedure to compute the temporal evolution of the distribution for a multiple-step forecast is presented. The evaluation of the proposed algorithm on three synthetic and two real data sets shows advantage over the compared baselines. | 0reject
|
Title: Direct Molecular Conformation Generation. Abstract: Molecular conformation generation, which is to generate 3 dimensional coordinates of all the atoms in a molecule, is an important task for bioinformatics and pharmacology. Most existing machine learning based methods first predict interatomic distances and then generate conformations based on them. This two-stage approach has a potential limitation that the predicted distances may conflict with each other, e.g., violating the triangle inequality. In this work, we propose a method that directly outputs the coordinates of atoms, so that there is no violation of constraints. The conformation generator of our method stacks multiple blocks, and each block outputs a conformation which is then refined by the following block. We adopt the variational auto-encoder (VAE) framework and use a latent variable to generate diverse conformations. To handle the roto-translation equivariance, we adopt a loss that is invariant to rotation and translation of molecule coordinates, by computing the minimal achievable distance after any rotation and translation. Our method outperforms strong baselines on four public datasets, which shows the effectiveness of our method and the great potential of the direct approach. The code is released at \url{https://github.com/DirectMolecularConfGen/DMCG}. | 0reject
|
Title: Explicit Recall for Efficient Exploration. Abstract: In this paper, we advocate the use of explicit memory for efficient exploration in reinforcement learning. This memory records structured trajectories that have led to interesting states in the past, and can be used by the agent to revisit those states more effectively. In high-dimensional decision making problems, where deep reinforcement learning is considered crucial, our approach provides a simple, transparent and effective way that can be naturally combined with complex, deep learning models. We show how such explicit memory may be used to enhance existing exploration algorithms such as intrinsically motivated ones and count-based ones, and demonstrate our method's advantages in various simulated environments. | 0reject
|
Title: Optimization and Adaptive Generalization of Three layer Neural Networks. Abstract: While there has been substantial recent work studying generalization of neural networks,
the ability of deep nets in automating the process of feature extraction still evades a thorough mathematical understanding.
As a step toward this goal, we analyze learning and generalization of a three-layer neural network with ReLU activations in a regime that goes beyond the linear approximation of the network, and is hence not captured by the common Neural Tangent Kernel. We show that despite nonconvexity of the empirical loss, a variant of SGD converges in polynomially many iterations to a good solution that generalizes. In particular, our generalization bounds are adaptive: they automatically optimize over a family of kernels that includes the Neural Tangent Kernel, to provide the tightest bound. | 1accept
|
Title: Closed-form Sample Probing for Learning Generative Models in Zero-shot Learning. Abstract: Generative model based approaches have led to significant advances in zero-shot learning (ZSL) over the past few years. These approaches typically aim to learn a conditional generator that synthesizes training samples of classes conditioned on class definitions. The final zero-shot learning model is then obtained by training a supervised classification model over the real and/or synthesized training samples of seen and unseen classes, combined. Therefore, naturally, the generative model needs to produce not only relevant samples, but also those that are sufficiently rich for classifier training purposes, which is handled by various heuristics in existing works. In this paper, we introduce a principled approach for training generative models {\em directly} for training data generation purposes. Our main observation is that the use of closed-form models opens doors to end-to-end training thanks to the differentiability of the solvers. In our approach, at each generative model update step, we fit a task-specific closed-form ZSL model from generated samples, and measure its loss on novel samples all within the compute graph, a procedure that we refer to as {\em sample probing}. In this manner, the generator receives feedback directly based on the value of its samples for model training purposes. Our experimental results show that the proposed sample probing approach improves the ZSL results even when integrated into state-of-the-art generative models.
| 1accept
|
Title: Holographic and other Point Set Distances for Machine Learning. Abstract: We introduce an analytic distance function for moderately sized point sets of known cardinality that is shown to have very desirable properties, both as a loss function as well as a regularizer for machine learning applications. We compare our novel construction to other point set distance functions and show proof of concept experiments for training neural networks end-to-end on point set prediction tasks such as object detection. | 0reject
|
Title: Learning and Data Selection in Big Datasets. Abstract: Finding a dataset of minimal cardinality to characterize the optimal parameters of a model is of paramount importance in machine learning and distributed optimization over a network. This paper investigates the compressibility of large datasets. More specifically, we propose a framework that jointly learns the input-output mapping as well as the most representative samples of the dataset (sufficient dataset). Our analytical results show that the cardinality of the sufficient dataset increases sub-linearly with respect to the original dataset size. Numerical evaluations of real datasets reveal a large compressibility, up to 95%, without a noticeable drop in the learnability performance, measured by the generalization error.
| 2withdrawn
|
Title: Graph Constrained Reinforcement Learning for Natural Language Action Spaces. Abstract: Interactive Fiction games are text-based simulations in which an agent interacts with the world purely through natural language. They are ideal environments for studying how to extend reinforcement learning agents to meet the challenges of natural language understanding, partial observability, and action generation in combinatorially-large text-based action spaces. We present KG-A2C, an agent that builds a dynamic knowledge graph while exploring and generates actions using a template-based action space. We contend that the dual uses of the knowledge graph to reason about game state and to constrain natural language generation are the keys to scalable exploration of combinatorially large natural language actions. Results across a wide variety of IF games show that KG-A2C outperforms current IF agents despite the exponential increase in action space size. | 1accept
|
Title: Modeling Winner-Take-All Competition in Sparse Binary Projections. Abstract: Inspired by the advances in biological science, the study of sparse binary projection models has attracted considerable recent research attention. The models project dense input samples into a higher-dimensional space and output sparse binary data representations after Winner-Take-All competition, subject to the constraint that the projection matrix is also sparse and binary. Following the work along this line, we developed a supervised-WTA model when training samples with both input and output representations are available, from which the optimal projection matrix can be obtained with a simple, efficient yet effective algorithm. We further extended the model and the algorithm to an unsupervised setting where only the input representation of the samples is available. In a series of empirical evaluation on similarity search tasks, the proposed models reported significantly improved results over the state-of-the-art methods in both search accuracy and running time. The successful results give us strong confidence that the work provides a highly practical tool to real world applications.
| 0reject
|
Title: Learning Algorithmic Solutions to Symbolic Planning Tasks with a Neural Computer. Abstract: A key feature of intelligent behavior is the ability to learn abstract strategies that transfer to unfamiliar problems. Therefore, we present a novel architecture, based on memory-augmented networks, that is inspired by the von Neumann and Harvard architectures of modern computers. This architecture enables the learning of abstract algorithmic solutions via Evolution Strategies in a reinforcement learning setting. Applied to Sokoban, sliding block puzzle and robotic manipulation tasks, we show that the architecture can learn algorithmic solutions with strong generalization and abstraction: scaling to arbitrary task configurations and complexities, and being independent of both the data representation and the task domain. | 0reject
|
Title: Data-driven Feature Sampling for Deep Hyperspectral Classification and Segmentation. Abstract: The high dimensionality of hyperspectral imaging forces unique challenges in scope, size and processing requirements. Motivated by the potential for an in-the-field cell sorting detector, we examine a Synechocystis sp. PCC 6803 dataset wherein cells are grown alternatively in nitrogen rich or deplete cultures. We use deep learning techniques to both successfully classify cells and generate a mask segmenting the cells/condition from the background. Further, we use the classification accuracy to guide a data-driven, iterative feature selection method, allowing the design neural networks requiring 90% fewer input features with little accuracy degradation. | 0reject
|
Title: AUGMENTED POLICY GRADIENT METHODS FOR EFFICIENT REINFORCEMENT LEARNING. Abstract: We propose a new mixture of model-based and model-free reinforcement learning
(RL) algorithms that combines the strengths of both RL methods. Our goal is to reduce the sample complexity of model-free approaches utilizing fictitious trajectory
rollouts performed on a learned dynamics model to improve the data efficiency of
policy gradient methods while maintaining the same asymptotic behaviour. We
suggest to use a special type of uncertainty quantification by a stochastic dynamics
model in which the next state prediction is randomly drawn from the distribution
predicted by the dynamics model. As a result, the negative effect of exploiting
erroneously optimistic regions in the dynamics model is addressed by next state
predictions based on an uncertainty aware ensemble of dynamics models. The
influence of the ensemble of dynamics models on the policy update is controlled
by adjusting the number of virtually performed rollouts in the next iteration according to the ratio of the real and virtual total reward. Our approach, which we
call Model-Based Policy Gradient Enrichment (MBPGE), is tested on a collection of benchmark tests including simulated robotic locomotion. We compare our
approach to plain model-free algorithms and a model-based one. Our evaluation
shows that MBPGE leads to higher learning rates in an early training stage and an
improved asymptotic behaviour. | 0reject
|
Title: Congested bandits: Optimal routing via short-term resets. Abstract: For traffic routing platforms, the choice of which route to recommend to a user depends on the congestion on these routes -- indeed, an individual's utility depends on the number of people using the recommended route at that instance. Motivated by this, we introduce the problem of Congested Bandits where each arm's reward is allowed to depend on the number of times it was played in the past $\Delta$ timesteps. This dependence on past history of actions leads to a dynamical system where an algorithm's present choices also affect its future pay-offs, and requires an algorithm to plan for this. We study the congestion aware formulation in the multi-armed bandit (MAB) setup and in the contextual bandit setup with linear rewards. For the multi-armed setup, we propose a UCB style algorithm and show that its policy regret scales as $\tilde{O}(\sqrt{K \Delta T})$. For the linear contextual bandit setup, our algorithm, based on an iterative least squares planner, achieves policy regret $\tilde{O}(\sqrt{dT} + \Delta)$. From an experimental standpoint, we corroborate the no-regret properties of our algorithms via a simulation study. | 0reject
|
Title: Incremental Learning on Growing Graphs. Abstract: Graphs have attracted numerous attention in varied areas and are dynamic in many scenarios. Among dynamic graphs, growing graphs with frequently expanding vertex and edge sets are typical and widely existed, e.g. the rapidly growing social networks. Confronting such growing data, existing methods on either static or dynamic graphs take the entire graph as a whole and may suffer from high computation cost and memory usage due to the continual growth of graphs. To tackle this problem, we introduce incremental graph learning (IGL), a general framework to formulate the learning on growing graphs in an incremental manner, where traditional graph learning method could be deployed as a basic model. We first analyze the problems of directly finetuning on the incremental part of graph, and theoretically discuss the unbiased and edge-preserved conditions of IGL. In our method, when the graph grows with new-coming data, we select or generate vertices and edges within restricted sizes from the previous graph to update current model together with the new data. Here, two strategies, i.e. sample-based and cluster-based, are proposed for learning with restricted time and space complexity. We conduct experiments on the node classification and link prediction tasks of multiple datasets. Experimental results and comparisons show that our method achieves satisfying performance with high efficiency on growing graphs. | 2withdrawn
|
Title: What are effective labels for augmented data? Improving robustness with AutoLabel. Abstract: A wide breadth of research has devised data augmentation approaches that can improve both accuracy and generalization performance for neural networks. However, augmented data can end up being far from the clean data and what is the appropriate label is less clear. Despite this, most existing work simply reuses the original label from the clean data, and the choice of label accompanying the augmented data is relatively less explored. In this paper, we propose AutoLabel to automatically learn the labels for augmented data, based on the distance between the clean distribution and augmented distribution. AutoLabel is built on label smoothing and is guided by the calibration-performance over a hold-out validation set. We show that AutoLabel is a generic framework that can be easily applied to existing data augmentation methods, including AugMix, mixup, and adversarial training. Experiments on CIFAR-10, CIFAR-100 and ImageNet show that AutoLabel can improve models' accuracy and calibration performance, especially under distributional shift. Additionally, we demonstrate that AutoLabel can help adversarial training by bridging the gap between clean accuracy and adversarial robustness. | 0reject
|
Title: BROS: A Pre-trained Language Model for Understanding Texts in Document. Abstract: Understanding document from their visual snapshots is an emerging and challenging problem that requires both advanced computer vision and NLP methods. Although the recent advance in OCR enables the accurate extraction of text segments, it is still challenging to extract key information from documents due to the diversity of layouts. To compensate for the difficulties, this paper introduces a pre-trained language model, BERT Relying On Spatiality (BROS), that represents and understands the semantics of spatially distributed texts. Different from previous pre-training methods on 1D text, BROS is pre-trained on large-scale semi-structured documents with a novel area-masking strategy while efficiently including the spatial layout information of input documents. Also, to generate structured outputs in various document understanding tasks, BROS utilizes a powerful graph-based decoder that can capture the relation between text segments. BROS achieves state-of-the-art results on four benchmark tasks: FUNSD, SROIE*, CORD, and SciTSR. Our experimental settings and implementation codes will be publicly available. | 0reject
|
Title: DSDF: Coordinated look-ahead strategy in stochastic multi-agent reinforcement learning. Abstract: Multi-Agent reinforcement learning has received lot of attention in recent years and have applications in many different areas. Existing methods involving Centralized Training and Decentralized execution, attempts to train the agents towards learning a pattern of coordinated actions to arrive at optimal joint policy. However if some agents are stochastic in their action to varying degrees, the above methods provides poor coordination among agents. In this paper we show how the stochasticity of agents, which could be a result of malfunction or aging of robots, can add to the uncertainty in coordination and thereby contribute to unsatisfactory global rewards. In such a scenario, the deterministic agents have to understand the behavior and limitations of the stochastic agents while the stochastic agents have to plan taking in cognizance their own limitations. Our proposed method, Deep Stochastic Discounted Factor (DSDF), tunes the discounted factor for the agents by using a learning representation of uncertainty to update the utility networks of individual agents. DSDF also helps in imparting an extent of reliability in coordination thereby granting stochastic agents tasks which are immediate and of shorter trajectory with deterministic ones taking the tasks which involve longer planning. Results on benchmark environments shows the efficacy of the proposed approach when compared with existing deterministic approaches. | 2withdrawn
|
Title: Thinking Deeper With Recurrent Networks: Logical Extrapolation Without Overthinking. Abstract: Classical machine learning systems perform best when they are trained and tested on the same distribution, and they lack a mechanism to increase model power after training is complete. In contrast, recent work has observed that recurrent networks can exhibit logical extrapolation; models trained only on small/simple problem instances can extend their abilities to solve large/complex instances at test time simply by performing more recurrent iterations. While preliminary results on these ``thinking systems'' are promising, existing recurrent systems, when iterated many times, often collapse rather than improve their performance. This ``overthinking'' phenomenon has prevented thinking systems from scaling to particularly large and complex problems. In this paper, we design a recall architecture that keeps an explicit copy of the problem instance in memory so that it cannot be forgotten. We also propose an incremental training routine that prevents the model from learning behaviors that are specific to iteration number and instead pushes it to learn behaviors that can be repeated indefinitely. Together, these design choices encourage models to converge to a steady state solution rather than deteriorate when many iterations are used. These innovations help to tackle the overthinking problem and boost deep thinking behavior on each of the benchmark tasks proposed by Schwarzschild et al. (2021a). | 0reject
|
Title: Understanding Intrinsic Robustness Using Label Uncertainty. Abstract: A fundamental question in adversarial machine learning is whether a robust classifier exists for a given task. A line of research has made some progress towards this goal by studying the concentration of measure, but we argue standard concentration fails to fully characterize the intrinsic robustness of a classification problem since it ignores data labels which are essential to any classification task. Building on a novel definition of label uncertainty, we empirically demonstrate that error regions induced by state-of-the-art models tend to have much higher label uncertainty than randomly-selected subsets. This observation motivates us to adapt a concentration estimation algorithm to account for label uncertainty, resulting in more accurate intrinsic robustness measures for benchmark image classification problems. | 1accept
|
Title: A Fine-Grained Analysis on Distribution Shift. Abstract: Robustness to distribution shifts is critical for deploying machine learning models in the real world. Despite this necessity, there has been little work in defining the underlying mechanisms that cause these shifts and evaluating the robustness of algorithms across multiple, different distribution shifts. To this end, we introduce a framework that enables fine-grained analysis of various distribution shifts. We provide a holistic analysis of current state-of-the-art methods by evaluating 19 distinct methods grouped into five categories across both synthetic and real-world datasets. Overall, we train more than 85K models. Our experimental framework can be easily extended to include new methods, shifts, and datasets. We find, unlike previous work (Gulrajani & Lopez-Paz, 2021), that progress has been made over a standard ERM baseline; in particular, pretraining and augmentations (learned or heuristic) offer large gains in many cases. However, the best methods are not consistent over different datasets and shifts. We will open source our experimental framework, allowing future work to evaluate new methods over multiple shifts to obtain a more complete picture of a method's effectiveness.
Code is available at github.com/deepmind/distribution_shift_framework.
| 1accept
|
Title: Two-Timescale Networks for Nonlinear Value Function Approximation. Abstract: A key component for many reinforcement learning agents is to learn a value function, either for policy evaluation or control. Many of the algorithms for learning values, however, are designed for linear function approximation---with a fixed basis or fixed representation. Though there have been a few sound extensions to nonlinear function approximation, such as nonlinear gradient temporal difference learning, these methods have largely not been adopted, eschewed in favour of simpler but not sound methods like temporal difference learning and Q-learning. In this work, we provide a two-timescale network (TTN) architecture that enables linear methods to be used to learn values, with a nonlinear representation learned at a slower timescale. The approach facilitates the use of algorithms developed for the linear setting, such as data-efficient least-squares methods, eligibility traces and the myriad of recently developed linear policy evaluation algorithms, to provide nonlinear value estimates. We prove convergence for TTNs, with particular care given to ensure convergence of the fast linear component under potentially dependent features provided by the learned representation. We empirically demonstrate the benefits of TTNs, compared to other nonlinear value function approximation algorithms, both for policy evaluation and control. | 1accept
|
Title: UAE-PUPET: An Uncertainty-Autoencoder-Based Privacy and Utility Preserving End-to-End Transformation. Abstract: We propose a new framework that deals with the privacy-utility tradeoff problem under two centralized settings: a dynamic setting and a constant setting. The dynamic setting corresponds to the min-max two-player game whereas the constant setting corresponds to a generator which tries to outperform an adversary already trained using ground truth data. In both settings, we use the same architecture consisting of a generator and a discriminator, where the generator consists of an encoder-decoder pair, and the discriminator consists of an adversary and a utility provider. Unlike previous research considering this kind of architecture, which leverage variational autoencoders (VAEs) based on learning a latent representation which is forced into a Gaussian assumption, our proposed technique removes the Gaussian assumption restriction on the latent variables, and only focuses on the end-to-end stochastic mapping of the input to privatized data. We also show that testing the privacy mechanism against a single adversary is usually not sufficient to capture the leakage of private information, as better adversaries can always be created by training under different conditions. Therefore, we test our proposed mechanism under five different types of adversary models. To compare privacy mechanisms under a fair framework, we propose a new metric called the Utility-Privacy Tradeoff (UPT) curve, obtained by using the upper convex hull of the utility-privacy tradeoff operation points achievable under the most powerful of the five adversary models. Finally, we test our framework on four different datasets: MNIST, Fashion MNIST, UCI Adult and US Census Demographic Data, providing a wide range of possible private and utility attributes. Through comparative analysis, our results show better privacy and utility guarantees, under our more rigorous adversary model, than the existing works, even when the latter are considered under their original restrictive single-adversary models. | 2withdrawn
|
Title: Manifold Distance Judge, an Adversarial Samples Defense Strategy Based on Service Orchestration. Abstract: Deep neural networks (DNNs) are playing an increasingly significant role in the modern world. However, they are weak to adversarial examples that are generated by adding specially crafted perturbations. Most defenses against adversarial examples focused on refining the DNN models, which often sacrifice the performance and computational cost of models on benign samples. In this paper, we propose a manifold distance detection method to distinguish between legitimate samples and adversarial samples by measuring the different distances on the manifold. The manifold distance detection method neither modifies the protected models nor requires knowledge of the process for generating adversarial samples. Inspired by the effectiveness of the manifold distance detection, we demonstrated a well-designed orchestrated defense strategy, named Manifold Distance Judge (MDJ), which selects the best image processing method that will effectively expand the manifold distance between legitimate and adversarial samples, and thus, enhances the performance of the following manifold distance detection method. Tests on the ImageNet dataset, the MDJ is effective against the most adversarial samples under whitebox, graybox, and blackbox attack scenarios. We show empirically that the orchestration strategy MDJ is significantly better than Feature Squeezing on the recall rate. Meanwhile, MDJ achieves high detection rates against CW attack and DI-FGSM attack. | 0reject
|
Title: FlowQA: Grasping Flow in History for Conversational Machine Comprehension. Abstract: Conversational machine comprehension requires a deep understanding of the conversation history. To enable traditional, single-turn models to encode the history comprehensively, we introduce Flow, a mechanism that can incorporate intermediate representations generated during the process of answering previous questions, through an alternating parallel processing structure. Compared to shallow approaches that concatenate previous questions/answers as input, Flow integrates the latent semantics of the conversation history more deeply. Our model, FlowQA, shows superior performance on two recently proposed conversational challenges (+7.2% F1 on CoQA and +4.0% on QuAC). The effectiveness of Flow also shows in other tasks. By reducing sequential instruction understanding to conversational machine comprehension, FlowQA outperforms the best models on all three domains in SCONE, with +1.8% to +4.4% improvement in accuracy. | 1accept
|
Title: Feature Attribution As Feature Selection. Abstract: Feature attribution methods identify "relevant" features as an explanation of a complex machine learning model. Several feature attribution methods have been proposed; however, only a few studies have attempted to define the "relevance" of each feature mathematically. In this study, we formalize the feature attribution problem as a feature selection problem. In our proposed formalization, there arise two possible definitions of relevance. We name the feature attribution problems based on these two relevances as Exclusive Feature Selection (EFS) and Inclusive Feature Selection (IFS). We show that several existing feature attribution methods can be interpreted as approximation algorithms for EFS and IFS. Moreover, through exhaustive experiments, we show that IFS is better suited as the formalization for the feature attribution problem than EFS. | 0reject
|
Title: Towards Finding Longer Proofs. Abstract: We present a reinforcement learning (RL) based guidance system for automated theorem proving geared towards Finding Longer Proofs (FLoP). FLoP is a step towards learning to reason by analogy, reducing the dependence on large scale search in automated theorem provers. We use several simple, structured datasets with very long proofs to show that FLoP can successfully generalise a single training proof to a large class of related problems, implementing a simple form of analogical reasoning. On these benchmarks, FLoP is competitive with strong theorem provers despite using very limited search. | 0reject
|
Title: Few-Shot Learning via Learning the Representation, Provably. Abstract: This paper studies few-shot learning via representation learning, where one uses $T$ source tasks with $n_1$ data per task to learn a representation in order to reduce the sample complexity of a target task for which there is only $n_2 (\ll n_1)$ data. Specifically, we focus on the setting where there exists a good common representation between source and target, and our goal is to understand how much a sample size reduction is possible. First, we study the setting where this common representation is low-dimensional and provide a risk bound of $\tilde{O}(\frac{dk}{n_1T} + \frac{k}{n_2})$ on the target task for the linear representation class; here $d$ is the ambient input dimension and $k (\ll d)$ is the dimension of the representation. This result bypasses the $\Omega(\frac{1}{T})$ barrier under the i.i.d. task assumption, and can capture the desired property that all $n_1T$ samples from source tasks can be \emph{pooled} together for representation learning. We further extend this result to handle a general representation function class and obtain a similar result. Next, we consider the setting where the common representation may be high-dimensional but is capacity-constrained (say in norm); here, we again demonstrate the advantage of representation learning in both high-dimensional linear regression and neural networks, and show that representation learning can fully utilize all $n_1T$ samples from source tasks. | 1accept
|
Title: Towards Achieving Adversarial Robustness Beyond Perceptual Limits. Abstract: The vulnerability of Deep Neural Networks to Adversarial Attacks has fuelled research towards building robust models. While most Adversarial Training algorithms aim towards defending attacks constrained within low magnitude $\ell_p$ norm bounds, real-world adversaries are not limited by such constraints. In this work, we aim to achieve adversarial robustness within larger bounds, against perturbations that may be perceptible, but do not change human (or Oracle) prediction. The presence of images that flip Oracle predictions and those that do not, makes this a challenging setting for adversarial robustness. We discuss the ideal goals of an adversarial defense algorithm beyond perceptual limits, and further highlight the shortcomings of naively extending existing training algorithms to higher perturbation bounds. In order to overcome these shortcomings, we propose a novel defense, Oracle-Aligned Adversarial Training (OA-AT), to align the predictions of the network with that of an Oracle during adversarial training. The proposed approach achieves state-of-the-art performance at large epsilon bounds (such as an $\ell_\infty$ bound of $16/255$ on CIFAR-10) while outperforming existing defenses (AWP, TRADES and PGD-AT) at standard perturbation bounds ($8/255$) as well. | 2withdrawn
|
Title: Provable Representation Learning for Imitation Learning via Bi-level Optimization. Abstract: A common strategy in modern learning systems is to learn a representation which is useful for many tasks, a.k.a, representation learning. We study this strategy in the imitation learning setting where multiple experts trajectories are available. We formulate representation learning as a bi-level optimization problem where the "outer" optimization tries to learn the joint representation and the "inner" optimization encodes the imitation learning setup and tries to learn task-specific parameters. We instantiate this framework for the cases where the imitation setting being behavior cloning and observation alone. Theoretically, we provably show using our framework that representation learning can reduce the sample complexity of imitation learning in both settings. We also provide proof-of-concept experiments to verify our theoretical findings. | 0reject
|
Title: Connecting the Dots Between MLE and RL for Sequence Generation. Abstract: Sequence generation models such as recurrent networks can be trained with a diverse set of learning algorithms. For example, maximum likelihood learning is simple and efficient, yet suffers from the exposure bias problem. Reinforcement learning like policy gradient addresses the problem but can have prohibitively poor exploration efficiency. A variety of other algorithms such as RAML, SPG, and data noising, have also been developed in different perspectives. This paper establishes a formal connection between these algorithms. We present a generalized entropy regularized policy optimization formulation, and show that the apparently divergent algorithms can all be reformulated as special instances of the framework, with the only difference being the configurations of reward function and a couple of hyperparameters. The unified interpretation offers a systematic view of the varying properties of exploration and learning efficiency. Besides, based on the framework, we present a new algorithm that dynamically interpolates among the existing algorithms for improved learning. Experiments on machine translation and text summarization demonstrate the superiority of the proposed algorithm. | 0reject
|
Title: Bandwidth-based Step-Sizes for Non-Convex Stochastic Optimization. Abstract: Many popular learning-rate schedules for deep neural networks combine a decaying trend with local perturbations that attempt to escape saddle points and bad local minima. We derive convergence guarantees for bandwidth-based step-sizes, a general class of learning-rates that are allowed to vary in a banded region. This framework includes many popular cyclic and non-monotonic step-sizes for which no theoretical guarantees were previously known. We provide worst-case guarantees for SGD on smooth non-convex problems under several bandwidth-based step sizes, including stagewise $1/\sqrt{t}$ and the popular \emph{step-decay} (``constant and then drop by a constant’’), which is also shown to be optimal. Moreover, we show that its momentum variant converges as fast as SGD with the bandwidth-based step-decay step-size. Finally, we propose novel step-size schemes in the bandwidth-based family and verify their efficiency on several deep neural network training tasks. | 0reject
|
Title: PI3NN: Out-of-distribution-aware Prediction Intervals from Three Neural Networks. Abstract: We propose a novel prediction interval (PI) method for uncertainty quantification, which addresses three major issues with the state-of-the-art PI methods. First, existing PI methods require retraining of neural networks (NNs) for every given confidence level and suffer from the crossing issue in calculating multiple PIs. Second, they usually rely on customized loss functions with extra sensitive hyperparameters for which fine tuning is required to achieve a well-calibrated PI. Third, they usually underestimate uncertainties of out-of-distribution (OOD) samples leading to over-confident PIs. Our PI3NN method calculates PIs from linear combinations of three NNs, each of which is independently trained using the standard mean squared error loss. The coefficients of the linear combinations are computed using root-finding algorithms to ensure tight PIs for a given confidence level. We theoretically prove that PI3NN can calculate PIs for a series of confidence levels without retraining NNs and it completely avoids the crossing issue. Additionally, PI3NN does not introduce any unusual hyperparameters resulting in a stable performance. Furthermore, we address OOD identification challenge by introducing an initialization scheme which provides reasonably larger PIs of the OOD samples than those of the in-distribution samples. Benchmark and real-world experiments show that our method outperforms several state-of-the-art approaches with respect to predictive uncertainty quality, robustness, and OOD samples identification. | 1accept
|
Title: On Multi-objective Policy Optimization as a Tool for Reinforcement Learning: Case Studies in Offline RL and Finetuning. Abstract: Many advances that have improved the robustness and efficiency of deep reinforcement learning (RL) algorithms can, in one way or another, be understood as introducing additional objectives or constraints in the policy optimization step. This includes ideas as far ranging as exploration bonuses, entropy regularization, and regularization toward teachers or data priors. Often, the task reward and auxiliary objectives are in conflict, and in this paper we argue that this makes it natural to treat these cases as instances of multi-objective (MO) optimization problems. We demonstrate how this perspective allows us to develop novel and more effective RL algorithms. In particular, we focus on offline RL and finetuning as case studies, and show that existing approaches can be understood as MO algorithms relying on linear scalarization. We hypothesize that replacing linear scalarization with a better algorithm can improve performance. We introduce Distillation of a Mixture of Experts (DiME), a new MORL algorithm that outperforms linear scalarization and can be applied to these non-standard MO problems. We demonstrate that for offline RL, DiME leads to a simple new algorithm that outperforms state-of-the-art. For finetuning, we derive new algorithms that learn to outperform the teacher policy. | 0reject
|
Title: Triple-Search: Differentiable Joint-Search of Networks, Precision, and Accelerators. Abstract: The record-breaking performance and prohibitive complexity of deep neural networks (DNNs) have ignited a substantial need for customized DNN accelerators which have the potential to boost DNN acceleration efficiency by orders-of-magnitude. While it has been recognized that maximizing DNNs' acceleration efficiency requires a joint design/search for three different yet highly coupled aspects, including the networks, adopted precision, and their accelerators, the challenges associated with such a joint search have not yet been fully discussed and addressed. First, to jointly search for a network and its precision via differentiable search, there exists a dilemma of whether to explode the memory consumption or achieve sub-optimal designs. Second, a generic and differentiable joint search of the networks and their accelerators is non-trivial due to (1) the discrete nature of the accelerator space and (2) the difficulty of obtaining operation-wise hardware cost penalties because some accelerator parameters are determined by the whole network. To this end, we propose a Triple-Search (TRIPS) framework to address the aforementioned challenges towards jointly searching for the network structure, precision, and accelerator in a differentiable manner, to efficiently and effectively explore the huge joint search space. Our TRIPS addresses the first challenge above via a heterogeneous sampling strategy to achieve unbiased search with constant memory consumption, and tackles the latter one using a novel co-search pipeline that integrates a generic differentiable accelerator search engine. Extensive experiments and ablation studies validate that both TRIPS generated networks and accelerators consistently outperform state-of-the-art (SOTA) designs (including co-search/exploration techniques, hardware-aware NAS methods, and DNN accelerators), in terms of search time, task accuracy, and accelerator efficiency. All codes will be released upon acceptance. | 0reject
|
Title: Sample Balancing for Improving Generalization under Distribution Shifts. Abstract: Deep neural networks have achieved striking performance when evaluated on testing data which share the same distribution with training ones, but can significantly fail otherwise. Therefore, eliminating the impact of distribution shifts between training and testing data is of paramount importance for building performance-promising deep models. Conventional methods (e.g. domain adaptation/generalization) assume either the availability of testing data or the known heterogeneity of training data (e.g. domain labels). In this paper, we consider a more challenging case where neither of the above information is available during the training phase. We propose to address this problem by removing the dependencies between features via reweighting training samples, which results in a more balanced distribution and helps deep models get rid of spurious correlations and, in turn, concentrate more on the true connection between features and labels. We conduct extensive experiments on object recognition benchmarks including PACS, VLCS, MNIST-M, and NICO which support the evaluation of generalization ability. The experimental results clearly demonstrate the effectiveness of the proposed method compared with state-of-the-art counterparts. | 2withdrawn
|
Title: Global-to-local Memory Pointer Networks for Task-Oriented Dialogue. Abstract: End-to-end task-oriented dialogue is challenging since knowledge bases are usually large, dynamic and hard to incorporate into a learning framework. We propose the global-to-local memory pointer (GLMP) networks to address this issue. In our model, a global memory encoder and a local memory decoder are proposed to share external knowledge. The encoder encodes dialogue history, modifies global contextual representation, and generates a global memory pointer. The decoder first generates a sketch response with unfilled slots. Next, it passes the global memory pointer to filter the external knowledge for relevant information, then instantiates the slots via the local memory pointers. We empirically show that our model can improve copy accuracy and mitigate the common out-of-vocabulary problem. As a result, GLMP is able to improve over the previous state-of-the-art models in both simulated bAbI Dialogue dataset and human-human Stanford Multi-domain Dialogue dataset on automatic and human evaluation. | 1accept
|
Title: Exploiting Redundancy: Separable Group Convolutional Networks on Lie Groups. Abstract: Group convolutional neural networks (G-CNNs) have been shown to increase parameter efficiency and model accuracy by incorporating geometric inductive biases. In this work, we investigate the properties of representations learned by regular G-CNNs, and show considerable parameter redundancy in group convolution kernels. This finding motivates further weight-tying by sharing convolution kernels over subgroups. To this end, we introduce convolution kernels that are separable over the subgroup and channel dimensions. In order to obtain equivariance to arbitrary affine Lie groups we provide a continuous parameterisation of separable convolution kernels. We evaluate our approach across several vision datasets, and show that our weight sharing leads to improved performance and computational efficiency. In many settings, separable G-CNNs outperform their non-separable counterpart, while only using a fraction of their training time. In addition, thanks to the increase in computational efficiency, we are able to implement G-CNNs equivariant to the $\mathrm{Sim(2)}$ group; the group of dilations, rotations and translations. $\mathrm{Sim(2)}$-equivariance further improves performance on all tasks considered. | 0reject
|
Title: Synthesizer: Rethinking Self-Attention for Transformer Models. Abstract: The dot product self-attention is known to be central and indispensable to state-of-the-art Transformer models. But is it really required? This paper investigates the true importance and contribution of the dot product-based self-attention mechanism on the performance of Transformer models. Via extensive experiments, we find that (1) random alignment matrices surprisingly perform quite competitively and (2) learning attention weights from token-token (query-key) interactions is useful but not that important after all. To this end, we propose \textsc{Synthesizer}, a model that learns synthetic attention weights without token-token interactions. In our experiments, we first show that simple Synthesizers achieve highly competitive performance when compared against vanilla Transformer models across a range of tasks, including machine translation, language modeling, text generation and GLUE/SuperGLUE benchmarks. When composed with dot product attention, we find that Synthesizers consistently outperform Transformers. Moreover, we conduct additional comparisons of Synthesizers against Dynamic Convolutions, showing that simple Random Synthesizer is not only $60\%$ faster but also improves perplexity by a relative $3.5\%$. Finally, we show that simple factorized Synthesizers can outperform Linformers on encoding only tasks. | 0reject
|
Title: Revisiting Layer-wise Sampling in Fast Training for Graph Convolutional Networks. Abstract: To accelerate the training of graph convolutional networks (GCN), many sampling-based methods have been developed for approximating the embedding aggregation. Among them, a layer-wise approach recursively performs importance sampling to select neighbors jointly for existing nodes in each layer. This paper revisits the approach from a matrix approximation perspective. We identify two issues in the existing layer-wise sampling methods: sub-optimal sampling probabilities and the approximation bias induced by sampling without replacement. We thus propose remedies to address these issues. The improvements are demonstrated by extensive analyses and experiments on common benchmarks. | 0reject
|
Title: CP-GAN: Towards a Better Global Landscape of GANs. Abstract: GANs have been very popular in data generation and unsupervised learning, but our understanding of GAN training is still very limited. One major reason is that GANs are often formulated as non-convex-concave min-max optimization. As a result, most recent studies focused on the analysis in the local region around the equilibrium. In this work, we perform a global analysis of GANs from two perspectives: the global landscape of the outer-optimization problem and the global behavior of the gradient descent dynamics. We find that the original GAN has exponentially many bad strict local minima which are perceived as mode-collapse, and the training dynamics (with linear discriminators) cannot escape mode collapse. To address these issues, we propose a simple modification to the original GAN, by coupling the generated samples and the true samples. We prove that the new formulation has no bad basins, and its training dynamics (with linear discriminators) has a Lyapunov function that leads to global convergence. Our experiments on standard datasets show that this simple loss outperforms the original GAN and WGAN-GP. | 0reject
|
Title: On Computation and Generalization of Generative Adversarial Imitation Learning. Abstract: Generative Adversarial Imitation Learning (GAIL) is a powerful and practical approach for learning sequential decision-making policies. Different from Reinforcement Learning (RL), GAIL takes advantage of demonstration data by experts (e.g., human), and learns both the policy and reward function of the unknown environment. Despite the significant empirical progresses, the theory behind GAIL is still largely unknown. The major difficulty comes from the underlying temporal dependency of the demonstration data and the minimax computational formulation of GAIL without convex-concave structure. To bridge such a gap between theory and practice, this paper investigates the theoretical properties of GAIL. Specifically, we show: (1) For GAIL with general reward parameterization, the generalization can be guaranteed as long as the class of the reward functions is properly controlled; (2) For GAIL, where the reward is parameterized as a reproducing kernel function, GAIL can be efficiently solved by stochastic first order optimization algorithms, which attain sublinear convergence to a stationary solution. To the best of our knowledge, these are the first results on statistical and computational guarantees of imitation learning with reward/policy function ap- proximation. Numerical experiments are provided to support our analysis.
| 1accept
|
Title: Representing Mixtures of Word Embeddings with Mixtures of Topic Embeddings. Abstract: A topic model is often formulated as a generative model that explains how each word of a document is generated given a set of topics and document-specific topic proportions. It is focused on capturing the word co-occurrences in a document and hence often suffers from poor performance in analyzing short documents. In addition, its parameter estimation often relies on approximate posterior inference that is either not scalable or suffering from large approximation error. This paper introduces a new topic-modeling framework where each document is viewed as a set of word embedding vectors and each topic is modeled as an embedding vector in the same embedding space. Embedding the words and topics in the same vector space, we define a method to measure the semantic difference between the embedding vectors of the words of a document and these of the topics, and optimize the topic embeddings to minimize the expected difference over all documents. Experiments on text analysis demonstrate that the proposed method, which is amenable to mini-batch stochastic gradient descent based optimization and hence scalable to big corpora, provides competitive performance in discovering more coherent and diverse topics and extracting better document representations. | 1accept
|
Title: Simulation of Human and Artificial Emotion (SHArE). Abstract: The framework for Simulation of Human and Artificial Emotion (SHArE) describes the architecture of emotion in terms of parameters transferable between psychology, neuroscience, and artificial intelligence. These parameters can be defined as abstract concepts, as with structural equation modeling or granularized down to the voltage levels of individual neurons. This model enables emotional trajectory design for humans which may lead to novel therapeutic solutions for various mental health concerns. For artificial intelligence, this work provides a compact notation which can be applied to neural networks as a means to observe the emotions and motivations of machines. | 0reject
|
Title: Mixture-of-Experts Variational Autoencoder for clustering and generating from similarity-based representations. Abstract: Clustering high-dimensional data, such as images or biological measurements, is a long-standing problem and has been studied extensively. Recently, Deep Clustering gained popularity due to the non-linearity of neural networks, which allows for flexibility in fitting the specific peculiarities of complex data. Here we introduce the Mixture-of-Experts Similarity Variational Autoencoder (MoE-Sim-VAE), a novel generative clustering model. The model can learn multi-modal distributions of high-dimensional data and use these to generate realistic data with high efficacy and efficiency. MoE-Sim-VAE is based on a Variational Autoencoder (VAE), where the decoder consists of a Mixture-of-Experts (MoE) architecture. This specific architecture allows for various modes of the data to be automatically learned by means of the experts. Additionally, we encourage the latent representation of our model to follow a Gaussian mixture distribution and to accurately represent the similarities between the data points. We assess the performance of our model on synthetic data, the MNIST benchmark data set, and a challenging real-world task of defining cell subpopulations from mass cytometry (CyTOF) measurements on hundreds of different datasets. MoE-Sim-VAE exhibits superior clustering performance on all these tasks in comparison to the baselines and we show that the MoE architecture in the decoder reduces the computational cost of sampling specific data modes with high fidelity. | 0reject
|
Title: Training GANs with Stronger Augmentations via Contrastive Discriminator. Abstract: Recent works in Generative Adversarial Networks (GANs) are actively revisiting various data augmentation techniques as an effective way to prevent discriminator overfitting. It is still unclear, however, that which augmentations could actually improve GANs, and in particular, how to apply a wider range of augmentations in training. In this paper, we propose a novel way to address these questions by incorporating a recent contrastive representation learning scheme into the GAN discriminator, coined ContraD. This "fusion" enables the discriminators to work with much stronger augmentations without increasing their training instability, thereby preventing the discriminator overfitting issue in GANs more effectively. Even better, we observe that the contrastive learning itself also benefits from our GAN training, i.e., by maintaining discriminative features between real and fake samples, suggesting a strong coherence between the two worlds: good contrastive representations are also good for GAN discriminators, and vice versa. Our experimental results show that GANs with ContraD consistently improve FID and IS compared to other recent techniques incorporating data augmentations, still maintaining highly discriminative features in the discriminator in terms of the linear evaluation. Finally, as a byproduct, we also show that our GANs trained in an unsupervised manner (without labels) can induce many conditional generative models via a simple latent sampling, leveraging the learned features of ContraD. Code is available at https://github.com/jh-jeong/ContraD. | 1accept
|
Title: Linear algebra with transformers. Abstract: Most applications of transformers to mathematics, from integration to theorem proving, focus on symbolic computation. In this paper, we show that transformers can be trained to perform numerical calculations with high accuracy. We consider problems of linear algebra: matrix transposition, addition, multiplication, eigenvalues and vectors, singular value decomposition, and inversion. Training small transformers (up to six layers) over datasets of random matrices, we achieve high accuracies (over 90%) on all problems. We also show that trained models can generalize out of their training distribution, and that out-of-domain accuracy can be greatly improved by working from more diverse datasets (in particular, by training from matrices with non-independent and identically distributed coefficients). Finally, we show that few-shot learning can be leveraged to retrain models to solve larger problems.
| 2withdrawn
|
Title: f-Domain-Adversarial Learning: Theory and Algorithms for Unsupervised Domain Adaptation with Neural Networks. Abstract: The problem of unsupervised domain adaptation arises in a variety of practical applications where the distribution of the training samples differs from those used at test time. The existing theory of domain adaptation derived generalization bounds based on divergence measures that are hard to optimize in practice. This has led to a large disconnect between theory and state-of-the-art methods. In this paper, we propose a novel domain-adversarial framework that introduces new theory for domain adaptation and leads to practical learning algorithms with neural networks. In particular, we derive a novel generalization bound that utilizes a new measure of discrepancy between distributions based on a variational characterization of f-divergences. We show that our bound recovers the theoretical results from Ben-David et al. (2010a) as a special case with a particular choice of divergence, and also supports divergences typically used in practice. We derive a general algorithm for domain-adversarial learning for the complete family of f-divergences. We provide empirical results for several f-divergences and show that some, not considered previously in domain-adversarial learning, achieve state-of-the-art results in practice. We provide empirical insights into how choosing a particular divergence affects the transfer performance on real-world datasets. By further recognizing the optimization problem as a Stackelberg game, we utilize the latest optimizers from the game optimization literature, achieving additional performance boosts in our training algorithm. We show that our f-domain adversarial framework achieves state-of-the-art results on the challenging Office-31 and Office-Home datasets without extra hyperparameters. | 0reject
|
Title: Alternating Multi-bit Quantization for Recurrent Neural Networks. Abstract: Recurrent neural networks have achieved excellent performance in many applications. However, on portable devices with limited resources, the models are often too large to deploy. For applications on the server with large scale concurrent requests, the latency during inference can also be very critical for costly computing resources. In this work, we address these problems by quantizing the network, both weights and activations, into multiple binary codes {-1,+1}. We formulate the quantization as an optimization problem. Under the key observation that once the quantization coefficients are fixed the binary codes can be derived efficiently by binary search tree, alternating minimization is then applied. We test the quantization for two well-known RNNs, i.e., long short term memory (LSTM) and gated recurrent unit (GRU), on the language models. Compared with the full-precision counter part, by 2-bit quantization we can achieve ~16x memory saving and ~6x real inference acceleration on CPUs, with only a reasonable loss in the accuracy. By 3-bit quantization, we can achieve almost no loss in the accuracy or even surpass the original model, with ~10.5x memory saving and ~3x real inference acceleration. Both results beat the exiting quantization works with large margins. We extend our alternating quantization to image classification tasks. In both RNNs and feedforward neural networks, the method also achieves excellent performance. | 1accept
|
Title: Multi-objective Optimization by Learning Space Partition. Abstract: In contrast to single-objective optimization (SOO), multi-objective optimization (MOO) requires an optimizer to find the Pareto frontier, a subset of feasible solutions that are not dominated by other feasible solutions. In this paper, we propose LaMOO, a novel multi-objective optimizer that learns a model from observed samples to partition the search space and then focus on promising regions that are likely to contain a subset of the Pareto frontier. The partitioning is based on the dominance number, which measures "how close'' a data point is to the Pareto frontier among existing samples. To account for possible partition errors due to limited samples and model mismatch, we leverage Monte Carlo Tree Search (MCTS) to exploit promising regions while exploring suboptimal regions that may turn out to contain good solutions later. Theoretically, we prove the efficacy of learning space partitioning via LaMOO under certain assumptions. Empirically, on the HyperVolume (HV) benchmark, a popular MOO metric, LaMOO substantially outperforms strong baselines on multiple real-world MOO tasks, by up to 225% in sample efficiency for neural architecture search on Nasbench201, and up to 10% for molecular design. | 1accept
|
Title: Learning Representation in Colour Conversion. Abstract: Colours can be represented in an infinite set of spaces highlighting distinct features. Here, we investigated the impact of colour spaces on the encoding capacity of a visual system that is subject to information compression, specifically variational autoencoders (VAEs) where bottlenecks are imposed. To this end, we propose a novel unsupervised task: colour space conversion (ColourConvNets). We trained several instances of VAEs whose input and output are in different colour spaces, e.g. from RGB to CIE L*a*b* (in total five colour spaces were examined). This allowed us to systematically study the influence of input-output colour spaces on the encoding efficiency and learnt representation. Our evaluations demonstrate that ColourConvNets with decorrelated output colour spaces produce higher quality images, also evident in pixel-wise low-level metrics such as colour difference ($\Delta E$), peak signal-to-noise ratio (PSNR) and structural similarity index measure (SSIM). We also assessed the ColourConvNets' capacity to reconstruct the global content in two downstream tasks: image classification (ImageNet) and scene segmentation (COCO). Our results show a 5-10% performance boost for decorrelating ColourConvNets with respect to the baseline network (whose input and output are RGB). Furthermore, we thoroughly analysed the finite embedding space of Vector Quantised VAEs with three different methods (single feature, hue shift and linear transformation). The interpretations reached with these techniques are in agreement suggesting that (i) luminance and chromatic information are encoded in separate embedding vectors, and (ii) the structure of the network's embedding space is determined by the output colour space. | 0reject
|
Title: Unsupervised Learning via Meta-Learning. Abstract: A central goal of unsupervised learning is to acquire representations from unlabeled data or experience that can be used for more effective learning of downstream tasks from modest amounts of labeled data. Many prior unsupervised learning works aim to do so by developing proxy objectives based on reconstruction, disentanglement, prediction, and other metrics. Instead, we develop an unsupervised meta-learning method that explicitly optimizes for the ability to learn a variety of tasks from small amounts of data. To do so, we construct tasks from unlabeled data in an automatic way and run meta-learning over the constructed tasks. Surprisingly, we find that, when integrated with meta-learning, relatively simple task construction mechanisms, such as clustering embeddings, lead to good performance on a variety of downstream, human-specified tasks. Our experiments across four image datasets indicate that our unsupervised meta-learning approach acquires a learning algorithm without any labeled data that is applicable to a wide range of downstream classification tasks, improving upon the embedding learned by four prior unsupervised learning methods. | 1accept
|
Title: A More Globally Accurate Dimensionality Reduction Method Using Triplets. Abstract: We first show that the commonly used dimensionality reduction (DR) methods such as t-SNE and LargeVis
poorly capture the global structure of the data in the low dimensional embedding. We show this via a number of tests for the DR methods that can be easily applied by any practitioner to the dataset at hand. Surprisingly enough, t-SNE performs the best w.r.t. the commonly used measures that reward the local neighborhood accuracy such as precision-recall while having the worst performance in our tests for global structure. We then contrast the performance of these two DR method
against our new method called TriMap. The main idea behind TriMap is to capture higher orders of structure with triplet information (instead of pairwise information used by t-SNE and LargeVis), and to minimize a robust loss function for satisfying the chosen triplets. We provide compelling experimental evidence on large natural datasets for the clear advantage of the TriMap DR results. As LargeVis, TriMap is fast and and provides comparable runtime on large datasets. | 0reject
|
Title: Improved Learning in Convolutional Neural Networks with Shifted Exponential Linear Units (ShELUs). Abstract: The Exponential Linear Unit (ELU) has been proven to speed up learning and improve the classification performance over activation functions such as ReLU and Leaky ReLU for convolutional neural networks. The reasons behind the improved behavior are that ELU reduces the bias shift, it saturates for large negative inputs and it is continuously differentiable. However, it remains open whether ELU has the optimal shape and we address the quest for a superior activation function.
We use a new formulation to tune a piecewise linear activation function during training, to investigate the above question, and learn the shape of the locally optimal activation function. With this tuned activation function, the classification performance is improved and the resulting, learned activation function shows to be ELU-shaped irrespective if it is initialized as a RELU, LReLU or ELU. Interestingly, the learned activation function does not exactly pass through the origin indicating that a shifted ELU-shaped activation function is preferable. This observation leads us to introduce the Shifted Exponential Linear Unit (ShELU) as a new activation function.
Experiments on Cifar-100 show that the classification performance is further improved when using the ShELU activation function in comparison with ELU. The improvement is achieved when learning an individual bias shift for each neuron. | 2withdrawn
|
Title: Lipschitz-constrained Unsupervised Skill Discovery. Abstract: We study the problem of unsupervised skill discovery, whose goal is to learn a set of diverse and useful skills with no external reward. There have been a number of skill discovery methods based on maximizing the mutual information (MI) between skills and states. However, we point out that their MI objectives usually prefer static skills to dynamic ones, which may hinder the application for downstream tasks. To address this issue, we propose Lipschitz-constrained Skill Discovery (LSD), which encourages the agent to discover more diverse, dynamic, and far-reaching skills. Another benefit of LSD is that its learned representation function can be utilized for solving goal-following downstream tasks even in a zero-shot manner — i.e., without further training or complex planning. Through experiments on various MuJoCo robotic locomotion and manipulation environments, we demonstrate that LSD outperforms previous approaches in terms of skill diversity, state space coverage, and performance on seven downstream tasks including the challenging task of following multiple goals on Humanoid. Our code and videos are available at https://shpark.me/projects/lsd/. | 1accept
|
Title: UNSUPERVISED CONVOLUTIONAL NEURAL NETWORKS FOR ACCURATE VIDEO FRAME INTERPOLATION WITH INTEGRATION OF MOTION COMPONENTS. Abstract: Optical flow and video frame interpolation are considered as a chicken-egg problem such that one problem affects the other and vice versa. This paper presents a deep neural network that integrates the flow network into the frame interpolation problem, with end-to-end learning. The proposed approach exploits the relationship between the two problems for quality enhancement of interpolation frames. Unlike recent convolutional neural networks, the proposed approach learns motions from natural video frames without graphical ground truth flows for training. This makes the network learn from extensive data and improve the performance. The motion information from the flow network guides interpolator networks to be trained to synthesize the interpolated frame accurately from motion scenarios. In addition, diverse datasets to cover various challenging cases that previous interpolations usually fail in is used for comparison. In all experimental datasets, the proposed network achieves better performance than state-of-art CNN based interpolations. With Middebury benchmark, compared with the top-ranked algorithm, the proposed network reduces an average interpolation error by about 9.3%. The proposed interpolation is ranked the 1st in Standard Deviation (SD) interpolation error, the 2nd in Average Interpolation Error among over 150 algorithms listed in the Middlebury interpolation benchmark. | 2withdrawn
|
Title: A Survey on Evidential Deep Learning For Single-Pass Uncertainty Estimation. Abstract: Popular approaches for quantifying predictive uncertainty in deep neural networks often involve a set of weights or models, for instance via ensembling or Monte Carlo Dropout. These techniques usually produce overhead by having to train multiple model instances or do not produce very diverse predictions. This survey aims to familiarize the reader with an alternative class of models based on the concept of Evidential Deep Learning: For unfamiliar data, they admit “what they don’t know” and fall back onto a prior belief. Furthermore, they allow uncertainty estimation in a single model and forward pass by parameterizing distributions over distributions. This survey recapitulates existing works, focusing on the implementation in a classification setting. Finally, we survey the application of the same paradigm to regression problems. We also provide a reflection on the strengths and weaknesses of the mentioned approaches compared to existing ones and provide the most central theoretical results in order to inform future research. | 0reject
|
Title: R5: Rule Discovery with Reinforced and Recurrent Relational Reasoning. Abstract: Systematicity, i.e., the ability to recombine known parts and rules to form new sequences while reasoning over relational data, is critical to machine intelligence. A model with strong systematicity is able to train on small-scale tasks and generalize to large-scale tasks. In this paper, we propose R5, a relational reasoning framework based on reinforcement learning that reasons over relational graph data and explicitly mines underlying compositional logical rules from observations. R5 has strong systematicity and being robust to noisy data. It consists of a policy value network equipped with Monte Carlo Tree Search to perform recurrent relational prediction and a backtrack rewriting mechanism for rule mining. By alternately applying the two components, R5 progressively learns a set of explicit rules from data and performs explainable and generalizable relation prediction. We conduct extensive evaluations on multiple datasets. Experimental results show that R5 outperforms various embedding-based and rule induction baselines on relation prediction tasks while achieving a high recall rate in discovering ground truth rules. | 1accept
|
Title: MGAN: Training Generative Adversarial Nets with Multiple Generators. Abstract: We propose in this paper a new approach to train the Generative Adversarial Nets (GANs) with a mixture of generators to overcome the mode collapsing problem. The main intuition is to employ multiple generators, instead of using a single one as in the original GAN. The idea is simple, yet proven to be extremely effective at covering diverse data modes, easily overcoming the mode collapsing problem and delivering state-of-the-art results. A minimax formulation was able to establish among a classifier, a discriminator, and a set of generators in a similar spirit with GAN. Generators create samples that are intended to come from the same distribution as the training data, whilst the discriminator determines whether samples are true data or generated by generators, and the classifier specifies which generator a sample comes from. The distinguishing feature is that internal samples are created from multiple generators, and then one of them will be randomly selected as final output similar to the mechanism of a probabilistic mixture model. We term our method Mixture Generative Adversarial Nets (MGAN). We develop theoretical analysis to prove that, at the equilibrium, the Jensen-Shannon divergence (JSD) between the mixture of generators’ distributions and the empirical data distribution is minimal, whilst the JSD among generators’ distributions is maximal, hence effectively avoiding the mode collapsing problem. By utilizing parameter sharing, our proposed model adds minimal computational cost to the standard GAN, and thus can also efficiently scale to large-scale datasets. We conduct extensive experiments on synthetic 2D data and natural image databases (CIFAR-10, STL-10 and ImageNet) to demonstrate the superior performance of our MGAN in achieving state-of-the-art Inception scores over latest baselines, generating diverse and appealing recognizable objects at different resolutions, and specializing in capturing different types of objects by the generators. | 1accept
|
Title: EVaDE : Event-Based Variational Thompson Sampling for Model-Based Reinforcement Learning. Abstract: Posterior Sampling for Reinforcement Learning (PSRL) is a well-known algorithm that augments model-based reinforcement learning (MBRL) algorithms with Thompson sampling. PSRL maintains posterior distributions of the environment transition dynamics and the reward function to procure posterior samples that are used to generate data for training the controller. Maintaining posterior distributions over all possible transition and reward functions for tasks with high dimensional state and action spaces is intractable. Recent works show that dropout used in conjunction with neural networks induce variational distributions that can approximate these posteriors. In this paper, we propose Event-based Variational Distributions for Exploration (EVaDE), variational distributions that are useful for MBRL, especially when the underlying domain is object-based. We leverage the general domain knowledge of object-based domains to design three types of event-based convolutional layers to direct exploration, namely the noisy event interaction layer, the noisy event weighting layer and the noisy event translation layer respectively. These layers rely on Gaussian dropouts and are inserted in between the layers of the deep neural network model to help facilitate variational Thompson sampling. We empirically show the effectiveness of EVaDE equipped Simulated Policy Learning (SimPLe) on a randomly selected suite of Atari games, where the number of agent environment interactions is limited to 100K. | 0reject
|
Title: FairFace: A Novel Face Attribute Dataset for Bias Measurement and Mitigation. Abstract: Existing public face image datasets are strongly biased toward Caucasian faces, and other races (e.g., Latino) are significantly underrepresented. The models trained from such datasets suffer from inconsistent classification accuracy, which limits the applicability of face analytic systems to non-White race groups. To mitigate the race bias problem in these datasets, we constructed a novel face image dataset containing 108,501 images which is balanced on race. We define 7 race groups: White, Black, Indian, East Asian, Southeast Asian, Middle Eastern, and Latino. Images were collected from the YFCC-100M Flickr dataset and labeled with race, gender, and age groups. Evaluations were performed on existing face attribute datasets as well as novel image datasets to measure the generalization performance. We find that the model trained from our dataset is substantially more accurate on novel datasets and the accuracy is consistent across race and gender groups. We also compare several commercial computer vision APIs and report their balanced accuracy across gender, race, and age groups. | 2withdrawn
|
Title: DivideMix: Learning with Noisy Labels as Semi-supervised Learning. Abstract: Deep neural networks are known to be annotation-hungry. Numerous efforts have been devoted to reducing the annotation cost when learning with deep networks. Two prominent directions include learning with noisy labels and semi-supervised learning by exploiting unlabeled data. In this work, we propose DivideMix, a novel framework for learning with noisy labels by leveraging semi-supervised learning techniques. In particular, DivideMix models the per-sample loss distribution with a mixture model to dynamically divide the training data into a labeled set with clean samples and an unlabeled set with noisy samples, and trains the model on both the labeled and unlabeled data in a semi-supervised manner. To avoid confirmation bias, we simultaneously train two diverged networks where each network uses the dataset division from the other network. During the semi-supervised training phase, we improve the MixMatch strategy by performing label co-refinement and label co-guessing on labeled and unlabeled samples, respectively. Experiments on multiple benchmark datasets demonstrate substantial improvements over state-of-the-art methods. Code is available at https://github.com/LiJunnan1992/DivideMix . | 1accept
|
Title: HalentNet: Multimodal Trajectory Forecasting with Hallucinative Intents. Abstract: Motion forecasting is essential for making intelligent decisions in robotic navigation. As a result, the multi-agent behavioral prediction has become a core component of modern human-robot interaction applications such as autonomous driving. Due to various intentions and interactions among agents, agent trajectories can have multiple possible futures. Hence, the motion forecasting model's ability to cover possible modes becomes essential to enable accurate prediction. Towards this goal, we introduce HalentNet to better model the future motion distribution in addition to a traditional trajectory regression learning objective by incorporating generative augmentation losses. We model intents with unsupervised discrete random variables whose training is guided by a collaboration between two key signals: A discriminative loss that encourages intents' diversity and a hallucinative loss that explores intent transitions (i.e., mixed intents) and encourages their smoothness. This regulates the neural network behavior to be more accurately predictive on uncertain scenarios due to the active yet careful exploration of possible future agent behavior. Our model's learned representation leads to better and more semantically meaningful coverage of the trajectory distribution. Our experiments show that our method can improve over the state-of-the-art trajectory forecasting benchmarks, including vehicles and pedestrians, for about 20% on average FDE and 50% on road boundary violation rate when predicting 6 seconds future. We also conducted human experiments to show that our predicted trajectories received 39.6% more votes than the runner-up approach and 32.2% more votes than our variant without hallucinative mixed intent loss. The code will be released soon. | 1accept
|
Title: Modeling Human Development: Effects of Blurred Vision on Category Learning in CNNs. Abstract: Recently, training convolutional neural networks (CNNs) using blurry images has been identified as a potential means to produce more robust models for facial recognition (Vogelsang et al. 2018). This method of training is intended to mimic biological visual development, as human visual acuity develops from near-blindness to normal acuity in the first three to four months of life (Kugelberg 1992). Object recognition develops in tandem during this time, and this developmental period has been shown to be critical for many visual tasks in later childhood and adulthood. We explore the effects of training CNNs on images with different levels of applied blur, including training regimens with progressively less blurry training sets. Using subsets of ImageNet (Russakovsky 2015), CNN performance is evaluated for both broad object recognition and fine-grained classification tasks. Results for AlexNet (Krizhevsky et al. 2012) and the more compact SqueezeNet (Iandola et al. 2016) are compared. Using blurry images for training on their own or as part of a training sequence increases classification accuracy across collections of images with different resolutions. At the same time, blurry training data causes little change to training convergence time and false positive classification certainty. Our findings support the utility of learning from sequences of blurry images for more robust image recognition. | 2withdrawn
|
Title: NASI: Label- and Data-agnostic Neural Architecture Search at Initialization. Abstract: Recent years have witnessed a surging interest in Neural Architecture Search (NAS). Various algorithms have been proposed to improve the search efficiency and effectiveness of NAS, i.e., to reduce the search cost and improve the generalization performance of the selected architectures, respectively. However, the search efficiency of these algorithms is severely limited by the need for model training during the search process. To overcome this limitation, we propose a novel NAS algorithm called NAS at Initialization (NASI) that exploits the capability of a Neural Tangent Kernel in being able to characterize the performance of candidate architectures at initialization, hence allowing model training to be completely avoided to boost the search efficiency. Besides the improved search efficiency, NASI also achieves competitive search effectiveness on various datasets like CIFAR-10/100 and ImageNet. Further, NASI is shown to be label- and data-agnostic under mild conditions, which guarantees the transferability of architectures selected by our NASI over different datasets. | 1accept
|
Title: Generalized Domain Adaptation with Covariate and Label Shift CO-ALignment. Abstract: Unsupervised knowledge transfer has a great potential to improve the generalizability of deep models to novel domains. Yet the current literature assumes that the label distribution is domain-invariant and only aligns the covariate or vice versa. In this paper, we explore the task of Generalized Domain Adaptation (GDA): How to transfer knowledge across different domains in the presence of both covariate and label shift? We propose a covariate and label distribution CO-ALignment (COAL) model to tackle this problem. Our model leverages prototype-based conditional alignment and label distribution estimation to diminish the covariate and label shifts, respectively. We demonstrate experimentally that when both types of shift exist in the data, COAL leads to state-of-the-art performance on several cross-domain benchmarks. | 0reject
|
Title: Attacking deep networks with surrogate-based adversarial black-box methods is easy. Abstract: A recent line of work on black-box adversarial attacks has revived the use of transfer from surrogate models by integrating it into query-based search. However, we find that existing approaches of this type underperform their potential, and can be overly complicated besides. Here, we provide a short and simple algorithm which achieves state-of-the-art results through a search which uses the surrogate network's class-score gradients, with no need for other priors or heuristics. The guiding assumption of the algorithm is that the studied networks are in a fundamental sense learning similar functions, and that a transfer attack from one to the other should thus be fairly "easy". This assumption is validated by the extremely low query counts and failure rates achieved: e.g. an untargeted attack on a VGG-16 ImageNet network using a ResNet-152 as the surrogate yields a median query count of 6 at a success rate of 99.9%. Code is available at https://github.com/fiveai/GFCS. | 1accept
|
Title: Multi-View Summarization and Activity Recognition Meet Edge Computing in IoT Environments. Abstract: Multi-view video summarization (MVS) lacks researchers’ attention due to their major challenges of inter-view correlations and overlapping of cameras. Most of the prior MVS works are offline, relying on only summary, needing extra communication bandwidth and transmission time with no focus on uncertain environments. Different from the existing methods, we propose edge intelligence based MVS and spatio-temporal features based activity recognition for IoT environments. We segment the multi-view videos on each slave device over edge into shots using light-weight CNN object detection model and compute mutual information among them to generate summary. Our system does not rely on summary only but encode and transmit it to a master device with neural computing stick (NCS) for intelligently computing inter-view correlations and efficiently recognizing activities, thereby saving computation resources, communication bandwidth, and transmission time. Experiments report an increase of 0.4 in F-measure score on MVS Office dataset as well as 0.2% and 2% increase in activity recognition accuracy over UCF-50 and YouTube 11 datasets, respectively, with lower storage and transmission time compared to state-of-the-art. The time complexity is decreased from 1.23 to 0.45 secs for a single frame processing, thereby generating 0.75 secs faster MVS. Furthermore, we made a new dataset by synthetically adding fog to an MVS dataset to show the adaptability of our system for both certain and uncertain surveillance environments. | 2withdrawn
|
Title: INFORMATION MAXIMIZATION AUTO-ENCODING. Abstract: We propose the Information Maximization Autoencoder (IMAE), an information theoretic approach to simultaneously learn continuous and discrete representations in an unsupervised setting. Unlike the Variational Autoencoder framework, IMAE starts from a stochastic encoder that seeks to map each input data to a hybrid discrete and continuous representation with the objective of maximizing the mutual information between the data and their representations. A decoder is included to approximate the posterior distribution of the data given their representations, where a high fidelity approximation can be achieved by leveraging the informative representations.
We show that the proposed objective is theoretically valid and provides a principled framework for understanding the tradeoffs regarding informativeness of each representation factor, disentanglement of representations, and decoding quality. | 0reject
|
Title: Graph Residual Flow for Molecular Graph Generation. Abstract: Statistical generative models for molecular graphs attract attention from many researchers from the fields of bio- and chemo-informatics. Among these models, invertible flow-based approaches are not fully explored yet. In this paper, we propose a powerful invertible flow for molecular graphs, called Graph Residual Flow (GRF). The GRF is based on residual flows, which are known for more flexible and complex non-linear mappings than traditional coupling flows. We theoretically derive non-trivial conditions such that GRF is invertible, and present a way of keeping the entire flows invertible throughout the training and sampling. Experimental results show that a generative model based on the proposed GRF achieve comparable generation performance, with much smaller number of trainable parameters compared to the existing flow-based model. | 0reject
|
Title: "Hey, that's not an ODE'": Faster ODE Adjoints with 12 Lines of Code. Abstract: Neural differential equations may be trained by backpropagating gradients via the adjoint method, which is another differential equation typically solved using an adaptive-step-size numerical differential equation solver. A proposed step is accepted if its error, \emph{relative to some norm}, is sufficiently small; else it is rejected, the step is shrunk, and the process is repeated. Here, we demonstrate that the particular structure of the adjoint equations makes the usual choices of norm (such as $L^2$) unnecessarily stringent. By replacing it with a more appropriate (semi)norm, fewer steps are unnecessarily rejected and the backpropagation is made faster. This requires only minor code modifications. Experiments on a wide range of tasks---including time series, generative modeling, and physical control---demonstrate a median improvement of 40% fewer function evaluations. On some problems we see as much as 62% fewer function evaluations, so that the overall training time is roughly halved. | 0reject
|
Title: Towards Counteracting Adversarial Perturbations to Resist Adversarial Examples. Abstract: Studies show that neural networks are susceptible to adversarial attacks. This exposes a potential threat to neural network-based artificial intelligence systems. We observe that the probability of the correct result outputted by the network increases by applying small perturbations generated for class labels other than the original predicted one to adversarial examples. Based on this observation, we propose a method of counteracting adversarial perturbations to resist adversarial examples. In our method, we randomly select a number of class labels and generate small perturbations for these selected labels. The generated perturbations are added together and then clamped to a specified space. The obtained perturbation is finally added to the adversarial example to counteract the adversarial perturbation contained in the example. The proposed method is applied at inference time and does not require retraining or finetuning the model. We validate the proposed method on CIFAR-10 and CIFAR-100. The experimental results demonstrate that our method effectively improves the defense performance of the baseline methods, especially against strong adversarial examples generated using more iterations. | 0reject
|
Title: Wyner VAE: A Variational Autoencoder with Succinct Common Representation Learning. Abstract: A new variational autoencoder (VAE) model is proposed that learns a succinct common representation of two correlated data variables for conditional and joint generation tasks. The proposed Wyner VAE model is based on two information theoretic problems---distributed simulation and channel synthesis---in which Wyner's common information arises as the fundamental limit of the succinctness of the common representation. The Wyner VAE decomposes a pair of correlated data variables into their common representation (e.g., a shared concept) and local representations that capture the remaining randomness (e.g., texture and style) in respective data variables by imposing the mutual information between the data variables and the common representation as a regularization term. The utility of the proposed approach is demonstrated through experiments for joint and conditional generation with and without style control using synthetic data and real images. Experimental results show that learning a succinct common representation achieves better generative performance and that the proposed model outperforms existing VAE variants and the variational information bottleneck method. | 0reject
|
Title: Q-learning with UCB Exploration is Sample Efficient for Infinite-Horizon MDP. Abstract: A fundamental question in reinforcement learning is whether model-free algorithms are sample efficient. Recently, Jin et al. (2018) proposed a Q-learning algorithm with UCB exploration policy, and proved it has nearly optimal regret bound for finite-horizon episodic MDP. In this paper, we adapt Q-learning with UCB-exploration bonus to infinite-horizon MDP with discounted rewards \emph{without} accessing a generative model. We show that the \textit{sample complexity of exploration} of our algorithm is bounded by $\tilde{O}({\frac{SA}{\epsilon^2(1-\gamma)^7}})$. This improves the previously best known result of $\tilde{O}({\frac{SA}{\epsilon^4(1-\gamma)^8}})$ in this setting achieved by delayed Q-learning (Strehlet al., 2006),, and matches the lower bound in terms of $\epsilon$ as well as $S$ and $A$ up to logarithmic factors. | 1accept
|
Title: A Benchmark for Voice-Face Cross-Modal Matching and Retrieval. Abstract: Cross-modal associations between a person's voice and face can be learned algorithmically, and this is a useful functionality in many audio and visual applications. The problem can be defined as two tasks: voice-face matching and retrieval. Recently, this topic has attracted much research attention, but it is still in its early stages of development, and evaluation protocols and test schemes need to be more standardized. Performance metrics for different subtasks are also scarce, and a benchmark for this problem needs to be established. In this paper, a baseline evaluation framework is proposed for voice-face matching and retrieval tasks. Test confidence is analyzed, and a confidence interval for estimated accuracy is proposed. Various state-of-the-art performances with high test confidence are achieved on a series of subtasks using the baseline method (called TriNet) included in this framework. The source code will be published along with the paper. The results of this study can provide a basis for future research on voice-face cross-modal learning. | 0reject
|
Title: Hierarchical Attention: What Really Counts in Various NLP Tasks. Abstract: Attention mechanisms in sequence to sequence models have shown great ability and wonderful performance in various natural language processing (NLP) tasks, such as sentence embedding, text generation, machine translation, machine reading comprehension, etc. Unfortunately, existing attention mechanisms only learn either high-level or low-level features. In this paper, we think that the lack of hierarchical mechanisms is a bottleneck in improving the performance of the attention mechanisms, and propose a novel Hierarchical Attention Mechanism (Ham) based on the weighted sum of different layers of a multi-level attention.
Ham achieves a state-of-the-art BLEU score of 0.26 on Chinese poem generation task and a nearly 6.5% averaged improvement compared with the existing machine reading comprehension models such as BIDAF and Match-LSTM. Furthermore, our experiments and theorems reveal that Ham has greater generalization and representation ability than existing attention mechanisms. | 0reject
|
Title: FSPool: Learning Set Representations with Featurewise Sort Pooling. Abstract: Traditional set prediction models can struggle with simple datasets due to an issue we call the responsibility problem. We introduce a pooling method for sets of feature vectors based on sorting features across elements of the set. This can be used to construct a permutation-equivariant auto-encoder that avoids this responsibility problem. On a toy dataset of polygons and a set version of MNIST, we show that such an auto-encoder produces considerably better reconstructions and representations. Replacing the pooling function in existing set encoders with FSPool improves accuracy and convergence speed on a variety of datasets. | 1accept
|
Title: AutoCoG: A Unified Data-Modal Co-Search Framework for Graph Neural Networks. Abstract: Neural architecture search (NAS) has demonstrated success in discovering promising architectures for vision or language modeling tasks, and it has recently been introduced to searching for graph neural networks (GNNs) as well. Despite the preliminary success, we argue that for GNNs, NAS has to be customized further, due to the topological complicacy of GNN input data (graph) as well as the notorious training instability. Besides optimizing the GNN model architecture, we propose to simultaneously optimize the input graph topology, via a set of parameterized data augmentation operators. That yields AutoCoG, the first unified data-model co-search NAS framework for GNNs. By defining a highly flexible data-model co-search space, AutoCoG is gracefully formulated as a principled bi-level optimization, that can be end-to-end solved by the differential search methods. Experiments demonstrate that AutoCoG produces state-of-the-art performance at standard benchmarks including Cora, PubMed, and Citeseer, outperforming both state-of-the-art hand-crafted GNNs as well as recent GNN-NAS methods. AutoCoG can also scale to searching deeper GCNs in larger-scale datasets. Our method consistently achieves state-of-the-art (SOTA) results on Cora, Citeseer, Pubmed, and ogbn-arxiv. Specifically, we achieve gains of up to 2.04% for Cora, 2.54% for Citeseer, 2.08% for Pubmed, and finally 0.83% for ogbn-arxiv on our benchmarks. | 2withdrawn
|
Title: Training Deep Spiking Neural Networks with Bio-plausible Learning Rules. Abstract: There exists a marked cleavage between the biological plausible approaches and the practical backpropagation-based approaches on how to train a deep spiking neural network (DSNN) with better performance. The well-known bio-plausible learning rule Spike-Timing-Dependent-Plasticity (STDP) cannot explain how the brain adjusts synaptic weights to encode information through accurate spike timing, while the widely applied backpropagation (BP) algorithms lack a biologically credible mechanism. In this work, we wish to answer the question of whether it is possible to train a DSNN using bio-plausible learning rules only and reach the comparable accuracy trained by the BP-based learning rules. We observed that the STDP learning rule calculated between the membrane potential waveform in the apical dendrite of a pyramidal cell and its input spike train with the help of local recurrent connections synthesized by somatostatin (SOM) interneurons is able to perform supervised learning. This architecture is also supported by recent observations of the brain's cortical microcircuits. This new view of how spiking neurons may accurately adjust their complex temporal dynamics with the help of special local feedback connections bridges the performance gap between the bio-plausible approaches and the BP-based approaches and provides a possible answer to how our brain learns. We verify our observation with a simplified spiking neuron model and two different cell types on several datasets and further provide theoretical proof of the equivalence between STDP and BP under a special circumstance. | 2withdrawn
|