Paper Group NANR 125
Understanding the (Un)interpretability of Natural Image Distributions Using Generative Models. MoET: Interpretable and Verifiable Reinforcement Learning via Mixture of Expert Trees. Universal Safeguarded Learned Convex Optimization with Guaranteed Convergence. Anomaly Detection Based on Unsupervised Disentangled Representation Learning in Combinati …
Understanding the (Un)interpretability of Natural Image Distributions Using Generative Models
| Title | Understanding the (Un)interpretability of Natural Image Distributions Using Generative Models |
| Authors | Anonymous |
| Abstract | Probability density estimation is a classical and well studied problem, but standard density estimation methods have historically lacked the power to model complex and high-dimensional image distributions. More recent generative models leverage the power of neural networks to implicitly learn and represent probability models over complex images. We describe methods to extract explicit probability density estimates from GANs, and explore the properties of these image density functions. We perform sanity check experiments to provide evidence that these probabilities are reasonable. However, we also show that density functions of natural images are difficult to interpret and thus limited in use. We study reasons for this lack of interpretability, and suggest that we can get better interpretability by doing density estimation on latent representations of images. |
| Tasks | Density Estimation |
| Published | 2020-01-01 |
| URL | https://openreview.net/forum?id=HJxw9lStPH |
| https://openreview.net/pdf?id=HJxw9lStPH | |
| PWC | https://paperswithcode.com/paper/understanding-the-uninterpretability-of-1 |
| Repo | |
| Framework | |
MoET: Interpretable and Verifiable Reinforcement Learning via Mixture of Expert Trees
| Title | MoET: Interpretable and Verifiable Reinforcement Learning via Mixture of Expert Trees |
| Authors | Anonymous |
| Abstract | Deep Reinforcement Learning (DRL) has led to many recent breakthroughs on complex control tasks, such as defeating the best human player in the game of Go. However, decisions made by the DRL agent are not explainable, hindering its applicability in safety-critical settings. Viper, a recently proposed technique, constructs a decision tree policy by mimicking the DRL agent. Decision trees are interpretable as each action made can be traced back to the decision rule path that lead to it. However, one global decision tree approximating the DRL policy has significant limitations with respect to the geometry of decision boundaries. We propose MoET, a more expressive, yet still interpretable model based on Mixture of Experts, consisting of a gating function that partitions the state space, and multiple decision tree experts that specialize on different partitions. We propose a training procedure to support non-differentiable decision tree experts and integrate it into imitation learning procedure of Viper. We evaluate our algorithm on four OpenAI gym environments, and show that the policy constructed in such a way is more performant and better mimics the DRL agent by lowering mispredictions and increasing the reward. We also show that MoET policies are amenable for verification using off-the-shelf automated theorem provers such as Z3. |
| Tasks | Game of Go, Imitation Learning |
| Published | 2020-01-01 |
| URL | https://openreview.net/forum?id=BJlxdCVKDB |
| https://openreview.net/pdf?id=BJlxdCVKDB | |
| PWC | https://paperswithcode.com/paper/moet-interpretable-and-verifiable-1 |
| Repo | |
| Framework | |
Universal Safeguarded Learned Convex Optimization with Guaranteed Convergence
| Title | Universal Safeguarded Learned Convex Optimization with Guaranteed Convergence |
| Authors | Anonymous |
| Abstract | Many applications require quickly and repeatedly solving a certain type of optimization problem, each time with new (but similar) data. However, state of the art general-purpose optimization methods may converge too slowly for real-time use. This shortcoming is addressed by “learning to optimize” (L2O) schemes, which construct neural networks from parameterized forms of the update operations of general-purpose methods. Inferences by each network form solution estimates, and networks are trained to optimize these estimates for a particular distribution of data. This results in task-specific algorithms (e.g., LISTA, ALISTA, and D-LADMM) that can converge order(s) of magnitude faster than general-purpose counterparts. We provide the first general L2O convergence theory by wrapping all L2O schemes for convex optimization within a single framework. Existing L2O schemes form special cases, and we give a practical guide for applying our L2O framework to other problems. Using safeguarding, our theory proves, as the number of network layers increases, the distance between inferences and the solution set goes to zero, i.e., each cluster point is a solution. Our numerical examples demonstrate the efficacy of our approach for both existing and new L2O methods. |
| Tasks | |
| Published | 2020-01-01 |
| URL | https://openreview.net/forum?id=HkxZigSYwS |
| https://openreview.net/pdf?id=HkxZigSYwS | |
| PWC | https://paperswithcode.com/paper/universal-safeguarded-learned-convex |
| Repo | |
| Framework | |
Anomaly Detection Based on Unsupervised Disentangled Representation Learning in Combination with Manifold Learning
| Title | Anomaly Detection Based on Unsupervised Disentangled Representation Learning in Combination with Manifold Learning |
| Authors | Anonymous |
| Abstract | Identifying anomalous samples from highly complex and unstructured data is a crucial but challenging task in a variety of intelligent systems. In this paper, we present a novel deep anomaly detection framework named AnoDM (standing for Anomaly detection based on unsupervised Disentangled representation learning and Manifold learning). The disentanglement learning is currently implemented by beta-VAE for automatically discovering interpretable factorized latent representations in a completely unsupervised manner. The manifold learning is realized by t-SNE for projecting the latent representations to a 2D map. We define a new anomaly score function by combining beta-VAE’s reconstruction error in the raw feature space and local density estimation in the t-SNE space. AnoDM was evaluated on both image and time-series data and achieved better results than models that use just one of the two measures and other deep learning methods. |
| Tasks | Anomaly Detection, Density Estimation, Representation Learning, Time Series |
| Published | 2020-01-01 |
| URL | https://openreview.net/forum?id=r1xHxgrKwr |
| https://openreview.net/pdf?id=r1xHxgrKwr | |
| PWC | https://paperswithcode.com/paper/anomaly-detection-based-on-unsupervised |
| Repo | |
| Framework | |
Evidence-Aware Entropy Decomposition For Active Deep Learning
| Title | Evidence-Aware Entropy Decomposition For Active Deep Learning |
| Authors | Anonymous |
| Abstract | We present a novel multi-source uncertainty prediction approach that enables deep learning (DL) models to be actively trained with much less labeled data. By leveraging the second-order uncertainty representation provided by subjective logic (SL), we conduct evidence-based theoretical analysis and formally decompose the predicted entropy over multiple classes into two distinct sources of uncertainty: vacuity and dissonance, caused by lack of evidence and conflict of strong evidence, respectively. The evidence based entropy decomposition provides deeper insights on the nature of uncertainty, which can help effectively explore a large and high-dimensional unlabeled data space. We develop a novel loss function that augments DL based evidence prediction with uncertainty anchor sample identification through kernel density estimation (KDE). The accurately estimated multiple sources of uncertainty are systematically integrated and dynamically balanced using a data sampling function for label-efficient active deep learning (ADL). Experiments conducted over both synthetic and real data and comparison with competitive AL methods demonstrate the effectiveness of the proposed ADL model. |
| Tasks | Density Estimation |
| Published | 2020-01-01 |
| URL | https://openreview.net/forum?id=B1lC62EKwr |
| https://openreview.net/pdf?id=B1lC62EKwr | |
| PWC | https://paperswithcode.com/paper/evidence-aware-entropy-decomposition-for |
| Repo | |
| Framework | |
From Variational to Deterministic Autoencoders
| Title | From Variational to Deterministic Autoencoders |
| Authors | Anonymous |
| Abstract | Variational Autoencoders (VAEs) provide a theoretically-backed and popular framework for deep generative models. However, learning a VAE from data poses still unanswered theoretical questions and considerable practical challenges. In this work, we propose an alternative framework for generative modeling that is simpler, easier to train, and deterministic, yet has many of the advantages of the VAE. We observe that sampling a stochastic encoder in a Gaussian VAE can be interpreted as simply injecting noise into the input of a deterministic decoder. We investigate how substituting this kind of stochasticity, with other explicit and implicit regularization schemes, can lead to an equally smooth and meaningful latent space without having to force it to conform to an arbitrarily chosen prior. To retrieve a generative mechanism to sample new data points, we introduce an ex-post density estimation step that can be readily applied to the proposed framework as well as existing VAEs, improving their sample quality. We show, in a rigorous empirical study, that the proposed regularized deterministic autoencoders are able to generate samples that are comparable to, or better than, those of VAEs and more powerful alternatives when applied to images as well as to structured data such as molecules. |
| Tasks | Density Estimation |
| Published | 2020-01-01 |
| URL | https://openreview.net/forum?id=S1g7tpEYDS |
| https://openreview.net/pdf?id=S1g7tpEYDS | |
| PWC | https://paperswithcode.com/paper/from-variational-to-deterministic-1 |
| Repo | |
| Framework | |
EXPLOITING SEMANTIC COHERENCE TO IMPROVE PREDICTION IN SATELLITE SCENE IMAGE ANALYSIS: APPLICATION TO DISEASE DENSITY ESTIMATION
| Title | EXPLOITING SEMANTIC COHERENCE TO IMPROVE PREDICTION IN SATELLITE SCENE IMAGE ANALYSIS: APPLICATION TO DISEASE DENSITY ESTIMATION |
| Authors | Rahman Sanya, Gilbert Maiga, Ernest Mwebaze |
| Abstract | High intra-class diversity and inter-class similarity is a characteristic of remote sensing scene image data sets currently posing significant difficulty for deep learning algorithms on classification tasks. To improve accuracy, post-classification methods have been proposed for smoothing results of model predictions. However, those approaches require an additional neural network to perform the smoothing operation, which adds overhead to the task. We propose an approach that involves learning deep features directly over neighboring scene images without requiring use of a cleanup model. Our approach utilizes a siamese network to improve the discriminative power of convolutional neural networks on a pair of neighboring scene images. It then exploits semantic coherence between this pair to enrich the feature vector of the image for which we want to predict a label. Empirical results show that this approach provides a viable alternative to existing methods. For example, our model improved prediction accuracy by 1 percentage point and dropped the mean squared error value by 0.02 over the baseline, on a disease density estimation task. These performance gains are comparable with results from existing post-classification methods, moreover without implementation overheads. |
| Tasks | Density Estimation |
| Published | 2020-01-01 |
| URL | https://openreview.net/forum?id=Skxn-JSYwr |
| https://openreview.net/pdf?id=Skxn-JSYwr | |
| PWC | https://paperswithcode.com/paper/exploiting-semantic-coherence-to-improve |
| Repo | |
| Framework | |
Instance adaptive adversarial training: Improved accuracy tradeoffs in neural nets
| Title | Instance adaptive adversarial training: Improved accuracy tradeoffs in neural nets |
| Authors | Anonymous |
| Abstract | Adversarial training is by far the most successful strategy for improving robustness of neural networks to adversarial attacks. Despite its success as a defense mechanism, adversarial training fails to generalize well to unperturbed test set. We hypothesize that this poor generalization is a consequence of adversarial training with uniform perturbation radius around every training sample. Samples close to decision boundary can be morphed into a different class under a small perturbation budget, and enforcing large margins around these samples produce poor decision boundaries that generalize poorly. Motivated by this hypothesis, we propose instance adaptive adversarial training – a technique that enforces sample-specific perturbation margins around every training sample. We show that using our approach, test accuracy on unperturbed samples improve with a marginal drop in robustness. Extensive experiments on CIFAR-10, CIFAR-100 and Imagenet datasets demonstrate the effectiveness of our proposed approach. |
| Tasks | Adversarial Defense |
| Published | 2020-01-01 |
| URL | https://openreview.net/forum?id=SyeOVTEFPH |
| https://openreview.net/pdf?id=SyeOVTEFPH | |
| PWC | https://paperswithcode.com/paper/instance-adaptive-adversarial-training-1 |
| Repo | |
| Framework | |
DUAL ADVERSARIAL MODEL FOR GENERATING 3D POINT CLOUD
| Title | DUAL ADVERSARIAL MODEL FOR GENERATING 3D POINT CLOUD |
| Authors | Anonymous |
| Abstract | Three-dimensional data, such as point clouds, are often composed of three coordinates with few featrues. In view of this, it is hard for common neural networks to learn and represent the characteristics directly. In this paper, we focus on latent space’s representation of data characteristics, introduce a novel generative framework based on AutoEncoder(AE) and Generative Adversarial Network(GAN) with extra well-designed loss. We embed this framework directly into the raw 3D-GAN, and experiments demonstrate the potential of the framework in regard of improving the performance on the public dataset compared with other point cloud generation models proposed in recent years. It even achieves state of-the-art performance. We also perform experiments on MNIST and exhibit an excellent result on 2D dataset. |
| Tasks | Point Cloud Generation |
| Published | 2020-01-01 |
| URL | https://openreview.net/forum?id=H1lyiaVFwB |
| https://openreview.net/pdf?id=H1lyiaVFwB | |
| PWC | https://paperswithcode.com/paper/dual-adversarial-model-for-generating-3d |
| Repo | |
| Framework | |
Improving Sample Efficiency in Model-Free Reinforcement Learning from Images
| Title | Improving Sample Efficiency in Model-Free Reinforcement Learning from Images |
| Authors | Anonymous |
| Abstract | Training an agent to solve control tasks directly from high-dimensional images with model-free reinforcement learning (RL) has proven difficult. The agent needs to learn a latent representation together with a control policy to perform the task. Fitting a high-capacity encoder using a scarce reward signal is not only extremely sample inefficient, but also prone to suboptimal convergence. Two ways to improve sample efficiency are to learn a good feature representation and use off-policy algorithms. We dissect various approaches of learning good latent features, and conclude that the image reconstruction loss is the essential ingredient that enables efficient and stable representation learning in image-based RL. Following these findings, we devise an off-policy actor-critic algorithm with an auxiliary decoder that trains end-to-end and matches state-of-the-art performance across both model-free and model-based algorithms on many challenging control tasks. We release our code to encourage future research on image-based RL. |
| Tasks | Image Reconstruction, Representation Learning |
| Published | 2020-01-01 |
| URL | https://openreview.net/forum?id=HklE01BYDB |
| https://openreview.net/pdf?id=HklE01BYDB | |
| PWC | https://paperswithcode.com/paper/improving-sample-efficiency-in-model-free |
| Repo | |
| Framework | |
Deep Interaction Processes for Time-Evolving Graphs
| Title | Deep Interaction Processes for Time-Evolving Graphs |
| Authors | Anonymous |
| Abstract | Time-evolving graphs are ubiquitous such as online transactions on an e-commerce platform and user interactions on social networks. While neural approaches have been proposed for graph modeling, most of them focus on static graphs. In this paper we present a principled deep neural approach that models continuous time-evolving graphs at multiple time resolutions based on a temporal point process framework. To model the dependency between latent dynamic representations of each node, we define a mixture of temporal cascades in which a node’s neural representation depends on not only this node’s previous representations but also the previous representations of related nodes that have interacted with this node. We generalize LSTM on this temporal cascade mixture and introduce novel time gates to model time intervals between interactions. Furthermore, we introduce a selection mechanism that gives important nodes large influence in both $k-$hop subgraphs of nodes in an interaction. To capture temporal dependency at multiple time-resolutions, we stack our neural representations in several layers and fuse them based on attention. Based on the temporal point process framework, our approach can naturally handle growth (and shrinkage) of graph nodes and interactions, making it inductive. Experimental results on interaction prediction and classification tasks – including a real-world financial application – illustrate the effectiveness of the time gate, the selection and attention mechanisms of our approach, as well as its superior performance over the alternative approaches. |
| Tasks | |
| Published | 2020-01-01 |
| URL | https://openreview.net/forum?id=HyxWteSFwS |
| https://openreview.net/pdf?id=HyxWteSFwS | |
| PWC | https://paperswithcode.com/paper/deep-interaction-processes-for-time-evolving |
| Repo | |
| Framework | |
Meta Reinforcement Learning with Autonomous Inference of Subtask Dependencies
| Title | Meta Reinforcement Learning with Autonomous Inference of Subtask Dependencies |
| Authors | Anonymous |
| Abstract | We propose and address a novel few-shot RL problem, where a task is characterized by a subtask graph which describes a set of subtasks and their dependencies that are unknown to the agent. The agent needs to quickly adapt to the task over few episodes during adaptation phase to maximize the return in the test phase. Instead of directly learning a meta-policy, we develop a Meta-learner with Subtask Graph Inference (MSGI), which infers the latent parameter of the task by interacting with the environment and maximizes the return given the latent parameter. To facilitate learning, we adopt an intrinsic reward inspired by upper confidence bound (UCB) that encourages efficient exploration. Our experiment results on two grid-world domains and StarCraft II environments show that the proposed method is able to accurately infer the latent task parameter, and to adapt more efficiently than existing meta RL and hierarchical RL methods. |
| Tasks | Efficient Exploration, Starcraft, Starcraft II |
| Published | 2020-01-01 |
| URL | https://openreview.net/forum?id=HkgsWxrtPB |
| https://openreview.net/pdf?id=HkgsWxrtPB | |
| PWC | https://paperswithcode.com/paper/meta-reinforcement-learning-with-autonomous |
| Repo | |
| Framework | |
Action Semantics Network: Considering the Effects of Actions in Multiagent Systems
| Title | Action Semantics Network: Considering the Effects of Actions in Multiagent Systems |
| Authors | Anonymous |
| Abstract | In multiagent systems (MASs), each agent makes individual decisions but all of them contribute globally to the system evolution. Learning in MASs is difficult since each agent’s selection of actions must take place in the presence of other co-learning agents. Moreover, the environmental stochasticity and uncertainties increase exponentially with the increase in the number of agents. Previous works borrow various multiagent coordination mechanisms into deep learning architecture to facilitate multiagent coordination. However, none of them explicitly consider action semantics between agents that different actions have different influence on other agents. In this paper, we propose a novel network architecture, named Action Semantics Network (ASN), that explicitly represents such action semantics between agents. ASN characterizes different actions’ influence on other agents using neural networks based on the action semantics between them. ASN can be easily combined with existing deep reinforcement learning (DRL) algorithms to boost their performance. Experimental results on StarCraft II and Neural MMO show ASN significantly improves the performance of state-of-the-art DRL approaches compared with several network architectures. |
| Tasks | Starcraft, Starcraft II |
| Published | 2020-01-01 |
| URL | https://openreview.net/forum?id=ryg48p4tPH |
| https://openreview.net/pdf?id=ryg48p4tPH | |
| PWC | https://paperswithcode.com/paper/action-semantics-network-considering-the-1 |
| Repo | |
| Framework | |
Trajectory growth through random deep ReLU networks
| Title | Trajectory growth through random deep ReLU networks |
| Authors | Anonymous |
| Abstract | This paper considers the growth in the length of one-dimensional trajectories as they are passed through deep ReLU neural networks, which, among other things, is one measure of the expressivity of deep networks. We generalise existing results, providing an alternative, simpler method for lower bounding expected trajectory growth through random networks, for a more general class of weights distributions, including sparsely connected networks. We illustrate this approach by deriving bounds for sparse-Gaussian, sparse-uniform, and sparse-discrete-valued random nets. We prove that trajectory growth can remain exponential in depth with these new distributions, including their sparse variants, with the sparsity parameter appearing in the base of the exponent. |
| Tasks | |
| Published | 2020-01-01 |
| URL | https://openreview.net/forum?id=rkedXgrKDH |
| https://openreview.net/pdf?id=rkedXgrKDH | |
| PWC | https://paperswithcode.com/paper/trajectory-growth-through-random-deep-relu |
| Repo | |
| Framework | |
Implicit λ-Jeffreys Autoencoders: Taking the Best of Both Worlds
| Title | Implicit λ-Jeffreys Autoencoders: Taking the Best of Both Worlds |
| Authors | Aibek Alanov, Max Kochurov, Artem Sobolev, Daniil Yashkov, Dmitry Vetrov |
| Abstract | We propose a new form of an autoencoding model which incorporates the best properties of variational autoencoders (VAE) and generative adversarial networks (GAN). It is known that GAN can produce very realistic samples while VAE does not suffer from mode collapsing problem. Our model optimizes λ-Jeffreys divergence between the model distribution and the true data distribution. We show that it takes the best properties of VAE and GAN objectives. It consists of two parts. One of these parts can be optimized by using the standard adversarial training, and the second one is the very objective of the VAE model. However, the straightforward way of substituting the VAE loss does not work well if we use an explicit likelihood such as Gaussian or Laplace which have limited flexibility in high dimensions and are unnatural for modelling images in the space of pixels. To tackle this problem we propose a novel approach to train the VAE model with an implicit likelihood by an adversarially trained discriminator. In an extensive set of experiments on CIFAR-10 and TinyImagent datasets, we show that our model achieves the state-of-the-art generation and reconstruction quality and demonstrate how we can balance between mode-seeking and mode-covering behaviour of our model by adjusting the weight λ in our objective. |
| Tasks | |
| Published | 2020-01-01 |
| URL | https://openreview.net/forum?id=Syxc1yrKvr |
| https://openreview.net/pdf?id=Syxc1yrKvr | |
| PWC | https://paperswithcode.com/paper/implicit-jeffreys-autoencoders-taking-the |
| Repo | |
| Framework | |
