Decentralized learning has been studied intensively in recent years motivated by its wide applications in the context of federated learning. The majority of previous research focuses on the offline setting in which the objective function is static. However, the offline setting becomes unrealistic in numerous machine learning applications that witness the change of massive data. In this paper, we propose \emph{decentralized online} algorithm for convex and continuous DR-submodular optimization, two classes of functions that are present in a variety of machine learning problems. Our algorithms achieve performance guarantees comparable to those in the centralized offline setting. Moreover, on average, each participant performs only a \emph{single} gradient computation per time step. Subsequently, we extend our algorithms to the bandit setting. Finally, we illustrate the competitive performance of our algorithms in real-world experiments.

Abstract--Multi-agent reinforcement learning (MARL) is a powerful technology to construct interactive artificial intelligent systems in various applications such as multi-robot control and self-driving cars. Unlike supervised model or single-agent reinforcement learning, which actively exploits network pruning, it is obscure that how pruning will work in multi-agent reinforcement learning with its cooperative and interactive characteristics. MARL, which are 7.13 higher and 12.43 more energy efficient Most importantly, the accelerator shows speedup up to 12.52 for MARL requires up to 942.9 GFLOPS for effective realtime In addition, as the MARL system is I. Current CPU and GPU-based systems cannot learning, known for solving long-term decision-making problems meet the above requirements due to the lack of computing effectively. It aims to train the action policy, which is units, high power consumption or low utilization for small about how an agent should take actions based on the feedback batch sizes. Instead, FPGA is emerging as a new solution for from the given environment to maximize cumulative rewards. For example, Recently, deep reinforcement learning (DRL) that utilizes a the Xilinx U280 acceleration card provides robust computing deep neural network (DNN) as an action policy has been proposed potential through 9,024 DSPs over 41MB of on-chip BRAM [1]-[4]. Although DRL stands out in various domains while showing less power consumption than GPU. In addition, such as industrial control and robotics [5]-[7], all of them the reconfigurability of FPGA allows the optimization of are limited to a single agent. Other significant applications irregular data access and parallelism with customized compact have started to employ interaction between multiple agents, for data format, where these hardware overhead occurs in network instance, analysis of language communication and the network pruning to handle computation-bound applications. Hence, extending DRL to have In this paper, we propose a FPGA-based acceleration system many agents is critical for developing intelligent systems named LearningGroup, to yield high performance for where agents can interact with each other or even with people.

Behavior prediction plays an important role in integrated autonomous driving software solutions. In behavior prediction research, interactive behavior prediction is a less-explored area, compared to single-agent behavior prediction. Predicting the motion of interactive agents requires initiating novel mechanisms to capture the joint behaviors of the interactive pairs. In this work, we formulate the end-to-end joint prediction problem as a sequential learning process of marginal learning and joint learning of vehicle behaviors. We propose ProspectNet, a joint learning block that adopts the weighted attention score to model the mutual influence between interactive agent pairs. The joint learning block first weighs the multi-modal predicted candidate trajectories, then updates the ego-agent's embedding via cross attention. Furthermore, we broadcast the individual future predictions for each interactive agent into a pair-wise scoring module to select the top $K$ prediction pairs. We show that ProspectNet outperforms the Cartesian product of two marginal predictions, and achieves comparable performance on the Waymo Interactive Motion Prediction benchmarks.

Resource-constrained mobile robots that lack the capability to be completely autonomous can rely on a human or AI supervisor acting at a remote site (e.g., control station or cloud) for their control. Such a supervised autonomy or cloud-based control of a robot poses high networking and computing capabilities requirements at both sites, which are not easy to achieve. This paper introduces and analyzes a new analog twin framework by synchronizing mobility between two mobile robots, where one robot acts as an analog twin to the other robot. We devise a novel priority-based supervised bilateral teleoperation strategy for goal navigation tasks to validate the proposed framework. The practical implementation of a supervised control strategy on this framework entails a mobile robot system divided into a Master-Client scheme over a communication channel where the Client robot resides on the site of operation guided by the Master robot through an agent (human or AI) from a remote location. The Master robot controls the Client robot with its autonomous navigation algorithm, which reacts to the predictive force received from the Client robot. We analyze the proposed strategy in terms of network performance (throughput and delay), task performance (tracking error and goal reach accuracy), and computing efficiency (memory and CPU utilization). Extensive simulations and real-world experiments demonstrate the method's novelty, flexibility, and versatility in realizing reactive planning applications with remote computational offloading capabilities compared to conventional offloading schemes.

Softmax policy gradient is a popular algorithm for policy optimization in single-agent reinforcement learning, particularly since projection is not needed for each gradient update. However, in multi-agent systems, the lack of central coordination introduces significant additional difficulties in the convergence analysis. Even for a stochastic game with identical interest, there can be multiple Nash Equilibria (NEs), which disables proof techniques that rely on the existence of a unique global optimum. Moreover, the softmax parameterization introduces non-NE policies with zero gradient, making it difficult for gradient-based algorithms in seeking NEs. In this paper, we study the finite time convergence of decentralized softmax gradient play in a special form of game, Markov Potential Games (MPGs), which includes the identical interest game as a special case. We investigate both gradient play and natural gradient play, with and without $\log$-barrier regularization. The established convergence rates for the unregularized cases contain a trajectory-dependent constant that can be arbitrarily large, whereas the $\log$-barrier regularization overcomes this drawback, with the cost of slightly worse dependence on other factors such as the action set size. An empirical study on an identical interest matrix game confirms the theoretical findings.

Abstract: In a continuous-time setting we investigate how the management of a firm controls a dynamic choice between two generic voluntary disclosure decision rules (strategies): one a full and transparent disclosure referred to as candid, the other, referred to as sparing, under which items only above a dynamic threshold value are disclosed. We show how management are rewarded with a reputational premium for being candid. The candid strategy is, however, costly because the alternative of sparing behaviour shields from a downgrade in disclosed low values. We show how parameters of the model such as news intensity, pay-for-performance and time-to-mandatory-disclosure determine the optimal choice of candid versus sparing strategies and optimal times for management to switch between the two. The private news updates received by management are modelled following a Poisson arrival process, occurring between the fixed (known) mandatory disclosure dates, such as fiscal years or quarters, with the news received by management generated by a background Black-Scholes model of economic activity and of its partial observation.

In this paper, we bridge this gap and Space-time graph neural networks (ST-GNNs) are recently developed study both signals and graphs that vary over time. We extend our architectures that learn efficient graph representations of timevarying previous work in ST-GNNs to accommodate time-varying graphs data. ST-GNNs are particularly useful in multi-agent systems, and prove their stability to stochastic graph perturbations. Our contributions due to their stability properties and their ability to respect communication can be summarized as follows: delays between the agents. In this paper we revisit the stability properties of ST-GNNs and prove that they are stable to (C1) We prove the stability of STGFs and ST-GNNs to stochastic stochastic graph perturbations. Our analysis suggests that ST-GNNs graph perturbations. Our result implies that ST-GNNs can are suitable for transfer learning on time-varying graphs and enables handle transfer learning to time-varying graphs.

This paper introduces a general multi-agent bandit model in which each agent is facing a finite set of arms and may communicate with other agents through a central controller in order to identify, in pure exploration, or play, in regret minimization, its optimal arm. The twist is that the optimal arm for each agent is the arm with largest expected mixed reward, where the mixed reward of an arm is a weighted sum of the rewards of this arm for all agents. This makes communication between agents often necessary. This general setting allows to recover and extend several recent models for collaborative bandit learning, including the recently proposed federated learning with personalization (Shi et al., 2021). In this paper, we provide new lower bounds on the sample complexity of pure exploration and on the regret. We then propose a near-optimal algorithm for pure exploration. This algorithm is based on phased elimination with two novel ingredients: a data-dependent sampling scheme within each phase, aimed at matching a relaxation of the lower bound.

Intrinsic rewards have been increasingly used to mitigate the sparse reward problem in single-agent reinforcement learning. These intrinsic rewards encourage the agent to look for novel experiences, guiding the agent to explore the environment sufficiently despite the lack of extrinsic rewards. Curiosity-driven exploration is a simple yet efficient approach that quantifies this novelty as the prediction error of the agent's curiosity module, an internal neural network that is trained to predict the agent's next state given its current state and action. We show here, however, that naively using this curiosity-driven approach to guide exploration in sparse reward cooperative multi-agent environments does not consistently lead to improved results. Straightforward multi-agent extensions of curiosity-driven exploration take into consideration either individual or collective novelty only and thus, they do not provide a distinct but collaborative intrinsic reward signal that is essential for learning in cooperative multi-agent tasks. In this work, we propose a curiosity-driven multi-agent exploration method that has the mixed objective of motivating the agents to explore the environment in ways that are individually and collectively novel. First, we develop a two-headed curiosity module that is trained to predict the corresponding agent's next observation in the first head and the next joint observation in the second head. Second, we design the intrinsic reward formula to be the sum of the individual and joint prediction errors of this curiosity module. We empirically show that the combination of our curiosity module architecture and intrinsic reward formulation guides multi-agent exploration more efficiently than baseline approaches, thereby providing the best performance boost to MARL algorithms in cooperative navigation environments with sparse rewards.

Large sequence model (SM) such as GPT series and BERT has displayed outstanding performance and generalization capabilities on vision, language, and recently reinforcement learning tasks. A natural follow-up question is how to abstract multi-agent decision making into an SM problem and benefit from the prosperous development of SMs. In this paper, we introduce a novel architecture named Multi-Agent Transformer (MAT) that effectively casts cooperative multi-agent reinforcement learning (MARL) into SM problems wherein the task is to map agents' observation sequence to agents' optimal action sequence. Our goal is to build the bridge between MARL and SMs so that the modeling power of modern sequence models can be unleashed for MARL. Central to our MAT is an encoder-decoder architecture which leverages the multi-agent advantage decomposition theorem to transform the joint policy search problem into a sequential decision making process; this renders only linear time complexity for multi-agent problems and, most importantly, endows MAT with monotonic performance improvement guarantee. Unlike prior arts such as Decision Transformer fit only pre-collected offline data, MAT is trained by online trials and errors from the environment in an on-policy fashion. To validate MAT, we conduct extensive experiments on StarCraftII, Multi-Agent MuJoCo, Dexterous Hands Manipulation, and Google Research Football benchmarks. Results demonstrate that MAT achieves superior performance and data efficiency compared to strong baselines including MAPPO and HAPPO. Furthermore, we demonstrate that MAT is an excellent few-short learner on unseen tasks regardless of changes in the number of agents. See our project page at https://sites.google.com/view/multi-agent-transformer.