Offline reinforcement learning (RL) is challenged by the distributional shift problem. To tackle this issue, existing works mainly focus on designing sophisticated policy constraints between the learned policy and the behavior policy. However, these constraints are applied equally to well-performing and inferior actions through uniform sampling, which might negatively affect the learned policy. In this article, we propose offline decoupled prioritized resampling (ODPR), which designs specialized priority functions for the suboptimal policy constraint issue in offline RL and employs unique decoupled resampling for training stability. Through theoretical analysis, we show that the distinctive priority functions induce a provable improved behavior policy by modifying the distribution of the original behavior policy, and when constrained to this improved policy, a policy-constrained offline RL algorithm is likely to yield a better solution. We provide two practical implementations to balance computation and performance: one estimates priorities based on a fit value network [advantage-based ODPR (ODPR-A)] and the other utilizes trajectory returns [return-based ODPR (ODPR-R)] for quick computation. As a highly compatible plug-and-play component, ODPR is evaluated with five prevalent offline RL algorithms: behavior cloning (BC), twin delayed deep deterministic policy gradient + BC (TD3 + BC), OnestepRL, conservative Q-learning (CQL), and implicit Q-learning (IQL). Our experiments confirm that both ODPR-A and ODPR-R significantly improve performance across all baseline methods. Moreover, ODPR-A can be effective in some challenging settings, i.e., without trajectory information. Code and pretrained weights are available at https://github.com/yueyang130/ODPR.

Qisen Yang, Zekun Wang, Honghui Chen, Shenzhi Wang, Yifan Pu, Xin Gao, Wenhao Huang, Shiji Song, Gao Huang. Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2024.

Offline reinforcement learning (RL) optimizes the policy on a previously collected dataset without any interactions with the environment, yet usually suffers from the distributional shift problem. To mitigate this issue, a typical solution is to impose a policy constraint on a policy improvement objective. However, existing methods generally adopt a "one-size-fits-all" practice, i.e., keeping only a single improvement-constraint balance for all the samples in a mini-batch or even the entire offline dataset. In this work, we argue that different samples should be treated with different policy constraint intensities. Based on this idea, a novel plug-in approach named guided offline RL (GORL) is proposed. GORL employs a guiding network, along with only a few expert demonstrations, to adaptively determine the relative importance of the policy improvement and policy constraint for every sample. We theoretically prove that the guidance provided by our method is rational and near-optimal. Extensive experiments on various environments suggest that GORL can be easily installed on most offline RL algorithms with statistically significant performance improvements.

The black-box nature of deep reinforcement learning (RL) hinders them from real-world applications. Therefore, interpreting and explaining RL agents have been active research topics in recent years. Existing methods for post-hoc explanations usually adopt the action matching principle to enable an easy understanding of vision-based RL agents. In this article, it is argued that the commonly used action matching principle is more like an explanation of deep neural networks (DNNs) than the interpretation of RL agents. It may lead to irrelevant or misplaced feature attribution when different DNNs’ outputs lead to the same rewards or different rewards result from the same outputs. Therefore, we propose to consider rewards, the essential objective of RL agents, as the essential objective of interpreting RL agents as well. To ensure reward consistency during interpretable feature discovery, a novel framework (RL interpreting RL, denoted as RL-in-RL) is proposed to solve the gradient disconnection from actions to rewards. We verify and evaluate our method on the Atari 2600 games as well as Duckietown, a challenging self-driving car simulator environment. The results show that our method manages to keep reward (or return) consistency and achieves high-quality feature attribution. Further, a series of analytical experiments validate our assumption of the action matching principle’s limitations.

We present iBall, a basketball video-watching system that leverages gaze-moderated embedded visualizations to facilitate game understanding and engagement of casual fans. Video broadcasting and online video platforms make watching basketball games increasingly accessible. Yet, for new or casual fans, watching basketball videos is often confusing due to their limited basketball knowledge and the lack of accessible, on-demand information to resolve their confusion. To assist casual fans in watching basketball videos, we compared the game-watching behaviors of casual and die-hard fans in a formative study and developed iBall based on the findings. iBall embeds visualizations into basketball videos using a computer vision pipeline, and automatically adapts the visualizations based on the game context and users’ gaze, helping casual fans appreciate basketball games without being overwhelmed. We confirmed the usefulness, usability, and engagement of iBall in a study with 16 casual fans, and further collected feedback from 8 die-hard fans.

Training practical agents usually involve offline and online reinforcement learning (RL) to balance the policy's performance and interaction costs. In particular, online fine-tuning has become a commonly used method to correct the erroneous estimates of out-of-distribution data learned in the offline training phase. However, even limited online interactions can be inaccessible or catastrophic for high-stake scenarios like healthcare and autonomous driving. In this work, we introduce an interaction-free training scheme dubbed Offline-with-Action-Preferences (OAP). The main insight is that, compared to online fine-tuning, querying the preferences between pre-collected and learned actions can be equally or even more helpful to the erroneous estimate problem. By adaptively encouraging or suppressing policy constraint according to action preferences, OAP could distinguish overestimation from beneficial policy improvement and thus attains a more accurate evaluation of unseen data. Theoretically, we prove a lower bound of the behavior policy's performance improvement brought by OAP. Moreover, comprehensive experiments on the D4RL benchmark and state-of-the-art algorithms demonstrate that OAP yields higher (29% on average) scores, especially on challenging AntMaze tasks (98% higher).

Offline-to-online reinforcement learning (RL) is a training paradigm that combines pre-training on a pre-collected dataset with fine-tuning in an online environment. However, the incorporation of online fine-tuning can intensify the well-known distributional shift problem. Existing solutions tackle this problem by imposing a policy constraint on the policy improvement objective in both offline and online learning. They typically advocate a single balance between policy improvement and constraints across diverse data collections. This one-size-fits-all manner may not optimally leverage each collected sample due to the significant variation in data quality across different states. To this end, we introduce Family Offline-to-Online RL (FamO2O), a simple yet effective framework that empowers existing algorithms to determine state-adaptive improvement-constraint balances. FamO2O utilizes a universal model to train a family of policies with different improvement/constraint intensities, and a balance model to select a suitable policy for each state. Theoretically, we prove that state-adaptive balances are necessary for achieving a higher policy performance upper bound. Empirically, extensive experiments show that FamO2O offers a statistically significant improvement over various existing methods, achieving state-of-the-art performance on the D4RL benchmark. Codes are available at https://github.com/LeapLabTHU/FamO2O.

Augmented sports videos, which combine visualizations and video effects to present data in actual scenes, can communicate insights engagingly and thus have been increasingly popular for sports enthusiasts around the world. Yet, creating augmented sports videos remains a challenging task, requiring considerable time and video editing skills. On the other hand, sports insights are often communicated using natural language, such as in commentaries, oral presentations, and articles, but usually lack visual cues. Thus, this work aims to facilitate the creation of augmented sports videos by enabling analysts to directly create visualizations embedded in videos using insights expressed in natural language. To achieve this goal, we propose a three-step approach - 1) detecting visualizable entities in the text, 2) mapping these entities into visualizations, and 3) scheduling these visualizations to play with the video - and analyzed 155 sports video clips and the accompanying commentaries for accomplishing these steps. Informed by our analysis, we have designed and implemented Sporthesia, a proof-of-concept system that takes racket-based sports videos and textual commentaries as the input and outputs augmented videos. We demonstrate Sporthesia's applicability in two exemplar scenarios, i.e., authoring augmented sports videos using text and augmenting historical sports videos based on auditory comments. A technical evaluation shows that Sporthesia achieves high accuracy (F1-score of 0.9) in detecting visualizable entities in the text. An expert evaluation with eight sports analysts suggests high utility, effectiveness, and satisfaction with our language-driven authoring method and provides insights for future improvement and opportunities.

Knowledge distillation is an effective approach to learn compact models (students) with the supervision of large and strong models (teachers). As empirically there exists a strong correlation between the performance of teacher and student models, it is commonly believed that a high performing teacher is preferred. Consequently, practitioners tend to use a well trained network or an ensemble of them as the teacher. In this paper, we observe that an intermediate model, i.e., a checkpoint in the middle of the training procedure, often serves as a better teacher compared to the fully converged model, although the former has much lower accuracy. More surprisingly, a weak snapshot ensemble of several intermediate models from a same training trajectory can outperform a strong ensemble of independently trained and fully converged models, when they are used as teachers. We show that this phenomenon can be partially explained by the information bottleneck principle: the feature representations of intermediate models can have higher mutual information regarding the input, and thus contain more ``dark knowledge'' for effective distillation. We further propose an optimal intermediate teacher selection algorithm based on maximizing the total task-related mutual information. Experiments verify its effectiveness and applicability. Our code is available at https://github.com/LeapLabTHU/CheckpointKD.

The backbone of traditional CNN classifier is generally considered as a feature extractor, followed by a linear layer which performs the classification. We propose a novel loss function, termed as CAM-loss, to constrain the embedded feature maps with the class activation maps (CAMs) which indicate the spatially discriminative regions of an image for particular categories. CAM-loss drives the backbone to express the features of target category and suppress the features of non-target categories or background, so as to obtain more discriminative feature representations. It can be simply applied in any CNN architecture with neglectable additional parameters and calculations. Experimental results show that CAM-loss is applicable to a variety of network structures and can be combined with mainstream regularization methods to improve the performance of image classification. The strong generalization ability of CAMloss is validated in the transfer learning and few shot learning tasks. Based on CAM-loss, we also propose a novel CAAM-CAM matching knowledge distillation method. This method directly uses the CAM generated by the teacher network to supervise the CAAM generated by the student network, which effectively improves the accuracy and convergence rate of the student network.