Hao Jin

Hao Jin, Yang Peng, Wenhao Yang et al.

We study a Federated Reinforcement Learning (FedRL) problem in which $n$ agents collaboratively learn a single policy without sharing the trajectories they collected during agent-environment interaction. We stress the constraint of environment heterogeneity, which means $n$ environments corresponding to these $n$ agents have different state transitions. To obtain a value function or a policy function which optimizes the overall performance in all environments, we propose two federated RL algorithms, \texttt{QAvg} and \texttt{PAvg}. We theoretically prove that these algorithms converge to suboptimal solutions, while such suboptimality depends on how heterogeneous these $n$ environments are. Moreover, we propose a heuristic that achieves personalization by embedding the $n$ environments into $n$ vectors. The personalization heuristic not only improves the training but also allows for better generalization to new environments.

Hao Jin, Liangyu Zhang, Zhihua Zhang

We study a Federated Reinforcement Learning (FedRL) problem with constraint heterogeneity. In our setting, we aim to solve a reinforcement learning problem with multiple constraints while $N$ training agents are located in $N$ different environments with limited access to the constraint signals and they are expected to collaboratively learn a policy satisfying all constraint signals. Such learning problems are prevalent in scenarios of Large Language Model (LLM) fine-tuning and healthcare applications. To solve the problem, we propose federated primal-dual policy optimization methods based on traditional policy gradient methods. Specifically, we introduce $N$ local Lagrange functions for agents to perform local policy updates, and these agents are then scheduled to periodically communicate on their local policies. Taking natural policy gradient (NPG) and proximal policy optimization (PPO) as policy optimization methods, we mainly focus on two instances of our algorithms, ie, {FedNPG} and {FedPPO}. We show that FedNPG achieves global convergence with an $\tilde{O}(1/\sqrt{T})$ rate, and FedPPO efficiently solves complicated learning tasks with the use of deep neural networks.

Hao Jin, Ce Shi, Tatsuhiro Tsuchiya

Context: Detecting arrays are mathematical structures aimed at fault identification in combinatorial interaction testing. However, they cannot be directly applied to systems that have constraints among test parameters. Such constraints are prevalent in real-world systems. Objectives: This paper proposes Constrained Detecting Arrays (CDAs), an extension of detecting arrays, which can be used for systems with constraints. Methods: The paper examines the properties and capabilities of CDAs with rigorous arguments. The paper also proposes two algorithms for constructing CDAs: One is aimed at generating minimum CDAs and the other is a heuristic algorithm aimed at fast generation of CDAs. The algorithms are evaluated through experiments using a benchmark dataset. Results: Experimental results show that the first algorithm can generate minimum CDAs if a sufficiently long generation time is allowed, and the second algorithm can generate minimum or near-minimum CDAs in a reasonable time. Conclusion: CDAs enhance detecting arrays to be applied to systems with constraints. The two proposed algorithms have different advantages with respect to the array size and generation time