Seizing Serendipity: Exploiting the Value of Past Success in Off-Policy Actor-Critic

Learning high-quality Q-value functions plays a key role in the success of many modern off-policy deep reinforcement learning (RL) algorithms. Previous works focus on addressing the value overestimation issue, an outcome of adopting function approximators and off-policy learning. Deviating from the common viewpoint, we observe that Q-values are indeed underestimated in the latter stage of the RL training process, primarily related to the use of inferior actions from the current policy in Bellman updates as compared to the more optimal action samples in the replay buffer. We hypothesize that this long-neglected phenomenon potentially hinders policy learning and reduces sample efficiency. Our insight to address this issue is to incorporate sufficient exploitation of past successes while maintaining exploration optimism. We propose the Blended Exploitation and Exploration (BEE) operator, a simple yet effective approach that updates Q-value using both historical best-performing actions and the current policy. The instantiations of our method in both model-free and model-based settings outperform state-of-the-art methods in various continuous control tasks and achieve strong performance in failure-prone scenarios and real-world robot tasks.

Subequivariant Graph Reinforcement Learning in 3D Environments

Learning a shared policy that guides the locomotion of different agents is of core interest in Reinforcement Learning (RL), which leads to the study of morphology-agnostic RL. However, existing benchmarks are highly restrictive in the choice of starting point and target point, constraining the movement of the agents within 2D space. In this work, we propose a novel setup for morphology-agnostic RL, dubbed Subequivariant Graph RL in 3D environments (3D-SGRL). Specifically, we first introduce a new set of more practical yet challenging benchmarks in 3D space that allows the agent to have full Degree-of-Freedoms to explore in arbitrary directions starting from arbitrary configurations. Moreover, to optimize the policy over the enlarged state-action space, we propose to inject geometric symmetry, i.e., subequivariance, into the modeling of the policy and Q-function such that the policy can generalize to all directions, improving exploration efficiency. This goal is achieved by a novel SubEquivariant Transformer (SET) that permits expressive message exchange. Finally, we evaluate the proposed method on the proposed benchmarks, where our method consistently and significantly outperforms existing approaches on single-task, multi-task, and zero-shot generalization scenarios. Extensive ablations are also conducted to verify our design. Code and videos are available on our project page: https://alpc91.github.io/SGRL/.

SE-Bridge: Speech Enhancement with Consistent Brownian Bridge

We propose SE-Bridge, a novel method for speech enhancement (SE). After recently applying the diffusion models to speech enhancement, we can achieve speech enhancement by solving a stochastic differential equation (SDE). Each SDE corresponds to a probabilistic flow ordinary differential equation (PF-ODE), and the trajectory of the PF-ODE solution consists of the speech states at different moments. Our approach is based on consistency model that ensure any speech states on the same PF-ODE trajectory, correspond to the same initial state. By integrating the Brownian Bridge process, the model is able to generate high-intelligibility speech samples without adversarial training. This is the first attempt that applies the consistency models to SE task, achieving state-of-the-art results in several metrics while saving 15 x the time required for sampling compared to the diffusion-based baseline. Our experiments on multiple datasets demonstrate the effectiveness of SE-Bridge in SE. Furthermore, we show through extensive experiments on downstream tasks, including Automatic Speech Recognition (ASR) and Speaker Verification (SV), that SE-Bridge can effectively support multiple downstream tasks.

Compacting Binary Neural Networks by Sparse Kernel Selection

Binary Neural Network (BNN) represents convolution weights with 1-bit values, which enhances the efficiency of storage and computation. This paper is motivated by a previously revealed phenomenon that the binary kernels in successful BNNs are nearly power-law distributed: their values are mostly clustered into a small number of codewords. This phenomenon encourages us to compact typical BNNs and obtain further close performance through learning non-repetitive kernels within a binary kernel subspace. Specifically, we regard the binarization process as kernel grouping in terms of a binary codebook, and our task lies in learning to select a smaller subset of codewords from the full codebook. We then leverage the Gumbel-Sinkhorn technique to approximate the codeword selection process, and develop the Permutation Straight-Through Estimator (PSTE) that is able to not only optimize the selection process end-to-end but also maintain the non-repetitive occupancy of selected codewords. Experiments verify that our method reduces both the model size and bit-wise computational costs, and achieves accuracy improvements compared with state-of-the-art BNNs under comparable budgets.

Intriguing Property of GAN for Remote Sensing Image Generation

Generative adversarial networks (GANs) have achieved remarkable progress in the natural image field. However, when applying GANs in the remote sensing (RS) image generation task, we discover an extraordinary phenomenon: the GAN model is more sensitive to the size of training data for RS image generation than for natural image generation. In other words, the generation quality of RS images will change significantly with the number of training categories or samples per category. In this paper, we first analyze this phenomenon from two kinds of toy experiments and conclude that the amount of feature information contained in the GAN model decreases with reduced training data. Based on this discovery, we propose two innovative adjustment schemes, namely Uniformity Regularization (UR) and Entropy Regularization (ER), to increase the information learned by the GAN model at the distributional and sample levels, respectively. We theoretically and empirically demonstrate the effectiveness and versatility of our methods. Extensive experiments on the NWPU-RESISC45 and PatternNet datasets show that our methods outperform the well-established models on RS image generation tasks.

A Survey on Causal Reinforcement Learning

While Reinforcement Learning (RL) achieves tremendous success in sequential decision-making problems of many domains, it still faces key challenges of data inefficiency and the lack of interpretability. Interestingly, many researchers have leveraged insights from the causality literature recently, bringing forth flourishing works to unify the merits of causality and address well the challenges from RL. As such, it is of great necessity and significance to collate these Causal Reinforcement Learning (CRL) works, offer a review of CRL methods, and investigate the potential functionality from causality toward RL. In particular, we divide existing CRL approaches into two categories according to whether their causality-based information is given in advance or not. We further analyze each category in terms of the formalization of different models, ranging from the Markov Decision Process (MDP), Partially Observed Markov Decision Process (POMDP), Multi-Arm Bandits (MAB), and Dynamic Treatment Regime (DTR). Moreover, we summarize the evaluation matrices and open sources while we discuss emerging applications, along with promising prospects for the future development of CRL.

Deep Reinforcement Learning for Robotic Pushing and Picking in Cluttered Environment

In this paper, a novel robotic grasping system is established to automatically pick up objects in cluttered scenes. A composite robotic hand composed of a suction cup and a gripper is designed for grasping the object stably. The suction cup is used for lifting the object from the clutter first and the gripper for grasping the object accordingly. We utilize the affordance map to provide pixel-wise lifting point candidates for the suction cup. To obtain a good affordance map, the active exploration mechanism is introduced to the system. An effective metric is designed to calculate the reward for the current affordance map, and a deep Q-Network (DQN) is employed to guide the robotic hand to actively explore the environment until the generated affordance map is suitable for grasping. Experimental results have demonstrated that the proposed robotic grasping system is able to greatly increase the success rate of the robotic grasping in cluttered scenes.

Introducing Expertise Logic into Graph Representation Learning from A Causal Perspective

Benefiting from the injection of human prior knowledge, graphs, as derived discrete data, are semantically dense so that models can efficiently learn the semantic information from such data. Accordingly, graph neural networks (GNNs) indeed achieve impressive success in various fields. Revisiting the GNN learning paradigms, we discover that the relationship between human expertise and the knowledge modeled by GNNs still confuses researchers. To this end, we introduce motivating experiments and derive an empirical observation that the human expertise is gradually learned by the GNNs in general domains. By further observing the ramifications of introducing expertise logic into graph representation learning, we conclude that leading the GNNs to learn human expertise can improve the model performance. By exploring the intrinsic mechanism behind such observations, we elaborate the Structural Causal Model for the graph representation learning paradigm. Following the theoretical guidance, we innovatively introduce the auxiliary causal logic learning paradigm to improve the model to learn the expertise logic causally related to the graph representation learning task. In practice, the counterfactual technique is further performed to tackle the insufficient training issue during optimization. Plentiful experiments on the crafted and real-world domains support the consistent effectiveness of the proposed method.

Tacchi: A Pluggable and Low Computational Cost Elastomer Deformation Simulator for Optical Tactile Sensors

Simulation is widely applied in robotics research to save time and resources. There have been several works to simulate optical tactile sensors that leverage either a smoothing method or Finite Element Method (FEM). However, elastomer deformation physics is not considered in the former method, whereas the latter requires a massive amount of computational resources like a computer cluster. In this work, we propose a pluggable and low computational cost simulator using the Taichi programming language for simulating optical tactile sensors, named as Tacchi . It reconstructs elastomer deformation using particles, which allows deformed elastomer surfaces to be rendered into tactile images and reveals contact information without suffering from high computational costs. Tacchi facilitates creating realistic tactile images in simulation, e.g., ones that capture wear-and-tear defects on object surfaces. In addition, the proposed Tacchi can be integrated with robotics simulators for a robot system simulation. Experiment results showed that Tacchi can produce images with better similarity to real images and achieved higher Sim2Real accuracy compared to the existing methods. Moreover, it can be connected with MuJoCo and Gazebo with only the requirement of 1G memory space in GPU compared to a computer cluster applied for FEM. With Tacchi, physical robot simulation with optical tactile sensors becomes possible. All the materials in this paper are available at https://github.com/zixichen007115/Tacchi .

Timestamp-Supervised Action Segmentation in the Perspective of Clustering

Video action segmentation aims to slice the video into several action segments. Recently, timestamp supervision has received much attention due to lower annotation costs. We find the frames near the boundaries of action segments are in the transition region between two consecutive actions and have unclear semantics, which we call ambiguous intervals. Most existing methods iteratively generate pseudo-labels for all frames in each video to train the segmentation model. However, ambiguous intervals are more likely to be assigned with noisy and incorrect pseudo-labels, which leads to performance degradation. We propose a novel framework to train the model under timestamp supervision including the following two parts. First, pseudo-label ensembling generates pseudo-label sequences with ambiguous intervals, where the frames have no pseudo-labels. Second, iterative clustering iteratively propagates the pseudo-labels to the ambiguous intervals by clustering, and thus updates the pseudo-label sequences to train the model. We further introduce a clustering loss, which encourages the features of frames within the same action segment more compact. Extensive experiments show the effectiveness of our method.