Supporting ultra-high data rates and flexible reconfigurability, Terahertz (THz) mesh networks are attractive for next-generation wireless backhaul systems that empower the integrated access and backhaul (IAB). In THz mesh backhaul networks, the efficient cross-layer routing and long-term resource allocation is yet an open problem due to dynamic traffic demands as well as possible link failures caused by the high directivity and high non-line-of-sight (NLoS) path loss of THz spectrum. In addition, unpredictable data traffic and the mixed integer programming property with the NP-hard nature further challenge the effective routing and long-term resource allocation design. In this paper, a deep reinforcement learning (DRL) based cross-layer design in THz mesh backhaul networks (DEFLECT) is proposed, by considering dynamic traffic demands and possible sudden link failures. In DEFLECT, a heuristic routing metric is first devised to facilitate resource efficiency (RE) enhancement regarding energy and sub-array usages. Furthermore, a DRL based resource allocation algorithm is developed to realize long-term RE maximization and fast recovery from broken links. Specifically in the DRL method, the exploited multi-task structure cooperatively benefits joint power and sub-array allocation. Additionally, the leveraged hierarchical architecture realizes tailored resource allocation for each base station and learned knowledge transfer for fast recovery. Simulation results show that DEFLECT routing consumes less resource, compared to the minimal hop-count metric. Moreover, unlike conventional DRL methods causing packet loss and second-level latency, DEFLECT DRL realizes the long-term RE maximization with no packet loss and millisecond-level latency, and recovers resource-efficient backhaul from broken links within 1s.

Constrained policy search (CPS) is a fundamental problem in offline reinforcement learning, which is generally solved by advantage weighted regression (AWR). However, previous methods may still encounter out-of-distribution actions due to the limited expressivity of Gaussian-based policies. On the other hand, directly applying the state-of-the-art models with distribution expression capabilities (i.e., diffusion models) in the AWR framework is insufficient since AWR requires exact policy probability densities, which is intractable in diffusion models. In this paper, we propose a novel approach called $\textbf{Diffusion Model based Constrained Policy Search (DiffCPS)}$, which tackles the diffusion-based constrained policy search without resorting to AWR. The theoretical analysis reveals our key insights by leveraging the action distribution of the diffusion model to eliminate the policy distribution constraint in the CPS and then utilizing the Evidence Lower Bound (ELBO) of diffusion-based policy to approximate the KL constraint. Consequently, DiffCPS admits the high expressivity of diffusion models while circumventing the cumbersome density calculation brought by AWR. Extensive experimental results based on the D4RL benchmark demonstrate the efficacy of our approach. We empirically show that DiffCPS achieves better or at least competitive performance compared to traditional AWR-based baselines as well as recent diffusion-based offline RL methods. The code is now available at $\href{https://github.com/felix-thu/DiffCPS}{https://github.com/felix-thu/DiffCPS}$.

In this effort, we consider the impact of regularization on the diversity of actions taken by policies generated from reinforcement learning agents trained using a policy gradient. Policy gradient agents are prone to entropy collapse, which means certain actions are seldomly, if ever, selected. We augment the optimization objective function for the policy with terms constructed from various $\varphi$-divergences and Maximum Mean Discrepancy which encourages current policies to follow different state visitation and/or action choice distribution than previously computed policies. We provide numerical experiments using MNIST, CIFAR10, and Spotify datasets. The results demonstrate the advantage of diversity-promoting policy regularization and that its use on gradient-based approaches have significantly improved performance on a variety of personalization tasks. Furthermore, numerical evidence is given to show that policy regularization increases performance without losing accuracy.

Policy gradient methods, where one searches for the policy of interest by maximizing the value functions using first-order information, become increasingly popular for sequential decision making in reinforcement learning, games, and control. Guaranteeing the global optimality of policy gradient methods, however, is highly nontrivial due to nonconcavity of the value functions. In this exposition, we highlight recent progresses in understanding and developing policy gradient methods with global convergence guarantees, putting an emphasis on their finite-time convergence rates with regard to salient problem parameters.

Zero-shot coordination (ZSC) is a new challenge focusing on generalizing learned coordination skills to unseen partners. Existing methods train the ego agent with partners from pre-trained or evolving populations. The agent's ZSC capability is typically evaluated with a few evaluation partners, including humans and agents, and reported by mean returns. Current evaluation methods for ZSC capability still need improvement in constructing diverse evaluation partners and comprehensively measuring ZSC capability. In this paper, we aim to create a reliable, comprehensive, and efficient evaluation method for ZSC capability. We formally define the ideal'diversity-complete' evaluation partners and propose the best response (BR) diversity, which is the population diversity of the BRs to the partners, to approximate the ideal evaluation partners. We propose an evaluation workflow including'diversity-complete' evaluation partners construction and a multidimensional metric, the Best Response Proximity (BR-Prox) metric. We re-evaluate strong ZSC methods in the Overcooked environment using the proposed evaluation workflow. Surprisingly, the results in some of the most used layouts fail to distinguish the performance of different ZSC methods. Moreover, the evaluated ZSC methods lack the ability to produce enough diverse and high-performing training partners. Our proposed evaluation workflow calls for a change in how we efficiently evaluate ZSC methods as a supplement to human evaluation. Zero-shot Coordination (ZSC) is a new challenge in training an agent named ego agent to have the capability to coordinate with unseen partners in cooperative AI (Hu et al., 2020).

This paper introduces a distributed, GPU-centric experience replay system, GEAR, designed to perform scalable reinforcement learning (RL) with large sequence models (such as transformers). With such models, existing systems such as Reverb face considerable bottlenecks in memory, computation, and communication. GEAR, however, optimizes memory efficiency by enabling the memory resources on GPU servers (including host memory and device memory) to manage trajectory data. Furthermore, it facilitates decentralized GPU devices to expedite various trajectory selection strategies, circumventing computational bottlenecks. GEAR is equipped with GPU kernels capable of collecting trajectories using zero-copy access to host memory, along with remote-directed-memory access over InfiniBand, improving communication efficiency. Cluster experiments have shown that GEAR can achieve performance levels up to 6x greater than Reverb when training state-of-the-art large RL models. GEAR is open-sourced at https://github.com/bigrl-team/gear.

The use of reinforcement learning (RL) in practical applications requires considering sub-optimal outcomes, which depend on the agent's familiarity with the uncertain environment. Dynamically adjusting the level of epistemic risk over the course of learning can tactically achieve reliable optimal policy in safety-critical environments and tackle the sub-optimality of a static risk level. In this work, we introduce a novel framework, Distributional RL with Online Risk Adaption (DRL-ORA), which can quantify the aleatory and epistemic uncertainties compositely and dynamically select the epistemic risk levels via solving a total variation minimization problem online. The risk level selection can be efficiently achieved through grid search using a Follow-The-Leader type algorithm, and its offline oracle is related to "satisficing measure" (in the decision analysis community) under a special modification of the loss function. We show multiple classes of tasks where DRL-ORA outperforms existing methods that rely on either a fixed risk level or manually predetermined risk level adaption. Given the simplicity of our modifications, we believe the framework can be easily incorporated into most RL algorithm variants.

Autonomous driving systems (ADSs) are capable of sensing the environment and making driving decisions autonomously. These systems are safety-critical, and testing them is one of the important approaches to ensure their safety. However, due to the inherent complexity of ADSs and the high dimensionality of their operating environment, the number of possible test scenarios for ADSs is infinite. Besides, the operating environment of ADSs is dynamic, continuously evolving, and full of uncertainties, which requires a testing approach adaptive to the environment. In addition, existing ADS testing techniques have limited effectiveness in ensuring the realism of test scenarios, especially the realism of weather conditions and their changes over time. Recently, reinforcement learning (RL) has demonstrated great potential in addressing challenging problems, especially those requiring constant adaptations to dynamic environments. To this end, we present DeepQTest, a novel ADS testing approach that uses RL to learn environment configurations with a high chance of revealing abnormal ADS behaviors. Specifically, DeepQTest employs Deep Q-Learning and adopts three safety and comfort measures to construct the reward functions. To ensure the realism of generated scenarios, DeepQTest defines a set of realistic constraints and introduces real-world weather conditions into the simulated environment. We employed three comparison baselines, i.e., random, greedy, and a state-of-the-art RL-based approach DeepCOllision, for evaluating DeepQTest on an industrial-scale ADS. Evaluation results show that DeepQTest demonstrated significantly better effectiveness in terms of generating scenarios leading to collisions and ensuring scenario realism compared with the baselines. In addition, among the three reward functions implemented in DeepQTest, Time-To-Collision is recommended as the best design according to our study.

Telemedicine applications have recently received substantial potential and interest, especially after the COVID-19 pandemic. Remote experience will help people get their complex surgery done or transfer knowledge to local surgeons, without the need to travel abroad. Even with breakthrough improvements in internet speeds, the delay in video streaming is still a hurdle in telemedicine applications. This imposes using image compression and region of interest (ROI) techniques to reduce the data size and transmission needs. This paper proposes a Deep Reinforcement Learning (DRL) model that intelligently adapts the ROI size and non-ROI quality depending on the estimated throughput. The delay and structural similarity index measure (SSIM) comparison are used to assess the DRL model. The comparison findings and the practical application reveal that DRL is capable of reducing the delay by 13% and keeping the overall quality in an acceptable range. Since the latency has been significantly reduced, these findings are a valuable enhancement to telemedicine applications.

Machine learning methods are progressively gaining acceptance in the electron microscopy community for de-noising, semantic segmentation, and dimensionality reduction of data post-acquisition. The introduction of the APIs by major instrument manufacturers now allows the deployment of ML workflows in microscopes, not only for data analytics but also for real-time decision-making and feedback for microscope operation. However, the number of use cases for real-time ML remains remarkably small. Here, we discuss some considerations in designing ML-based active experiments and pose that the likely strategy for the next several years will be human-in-the-loop automated experiments (hAE). In this paradigm, the ML learning agent directly controls beam position and image and spectroscopy acquisition functions, and human operator monitors experiment progression in real-and feature space of the system and tunes the policies of the ML agent to steer the experiment towards specific objectives. One of the hallmarks of the meeting was the large number of presentations on machine learning (ML) in microscopy, ranging from denoising, unsupervised data analysis via variational autoencoders, and supervised learning applications for semantic segmentations and feature identification. Remarkably, by now most manufacturers offer or have plans to offer Python application programming interfaces (APIs), allowing the deployment of the codes on operational microscopes. From this perspective, the technical barriers for the broad implementation of automated microscopy in which ML algorithms analyze the data streaming from instrument detectors and make decisions based on this data are lower than ever.