AAMAS’21: Difference Rewards Policy Gradients

At next AAMAS, Jacopo Castellini, Sam Devlin, Rahul Savani and myself, will present our work on combining difference rewards and policy gradient methods.

Main idea: for differencing the function needs to be quite accurate. As such doing differencing on Q-functions (as COMA) might not be ideal. We instead perform the differencing on the reward function, which may be known and otherwise easier to learn (stationary). Our results show potential for great improvements especially for larger number of agents.

Are Multiple Agents the Solution, and not the Problem, to Non-Stationarity?

That is what we explore in our AAMAS’21 blue sky paper.

The idea is to explicitly model non-stationarity as part of an environmental shift game (ESG). This enables us to predict and even steer the shifts that would occur, while dealing with epistemic uncertainty in a robust manner.

NeurIPS Camready: Influence-Augmented Online Planning

The camready version of Influence-Augmented Online Planning for Complex Environments is now available.

In this work, we show that by learning approximate representations of influence, we can speed up online planning (POMCP) sufficiently to get better performance when the time for online decision making is constrained.

AAMAS: Maximizing Information Gain via Prediction Rewards

This paper tackles the problem of active perception: taking actions to minimize one’s uncertainty. It further formalizes the link between information gain and prediction rewards, and uses this to propose a deep-learning approach to optimize active perception from a data set, thus obviating the need for a complex POMDP model.