Deep Interactive Bayesian Reinforcement Learning via Meta-Learning
This work tackles the problem of efficient adaptation to unknown opponent strategies in multi-agent systems, which is a significant challenge for developing more robust and intelligent AI agents. It represents an incremental improvement in handling the computational complexity of IBRL.
This paper addresses the intractability of Interactive Bayesian Reinforcement Learning (IBRL) when agents need to learn about and adapt to unknown strategies of other agents. The authors propose a meta-learning approach that combines sequential and hierarchical Variational Auto-Encoders to approximate belief inference and Bayes-optimal behavior. Their method empirically outperforms existing model-free, memory-free, and structure-agnostic approaches.
Agents that interact with other agents often do not know a priori what the other agents' strategies are, but have to maximise their own online return while interacting with and learning about others. The optimal adaptive behaviour under uncertainty over the other agents' strategies w.r.t. some prior can in principle be computed using the Interactive Bayesian Reinforcement Learning framework. Unfortunately, doing so is intractable in most settings, and existing approximation methods are restricted to small tasks. To overcome this, we propose to meta-learn approximate belief inference and Bayes-optimal behaviour for a given prior. To model beliefs over other agents, we combine sequential and hierarchical Variational Auto-Encoders, and meta-train this inference model alongside the policy. We show empirically that our approach outperforms existing methods that use a model-free approach, sample from the approximate posterior, maintain memory-free models of others, or do not fully utilise the known structure of the environment.