BRExIt: On Opponent Modelling in Expert Iteration
This work addresses multi-agent learning for game theory, offering an incremental improvement over existing methods.
The paper tackles the problem of accelerating learning in games by incorporating opponent models into the Expert Iteration algorithm, resulting in BRExIt, which learns better performing policies than ExIt as shown in empirical tests.
Finding a best response policy is a central objective in game theory and multi-agent learning, with modern population-based training approaches employing reinforcement learning algorithms as best-response oracles to improve play against candidate opponents (typically previously learnt policies). We propose Best Response Expert Iteration (BRExIt), which accelerates learning in games by incorporating opponent models into the state-of-the-art learning algorithm Expert Iteration (ExIt). BRExIt aims to (1) improve feature shaping in the apprentice, with a policy head predicting opponent policies as an auxiliary task, and (2) bias opponent moves in planning towards the given or learnt opponent model, to generate apprentice targets that better approximate a best response. In an empirical ablation on BRExIt's algorithmic variants against a set of fixed test agents, we provide statistical evidence that BRExIt learns better performing policies than ExIt.