Summarising and Comparing Agent Dynamics with Contrastive Spatiotemporal Abstraction
This provides a complementary tool for agent interpretability, addressing the need for better understanding of learning processes in control agents.
The paper tackles the problem of summarizing and contrasting agent dynamics in evolving systems like reinforcement learning, by introducing a data-driven, model-agnostic technique that aggregates transition data using information-theoretic divergence, resulting in human-interpretable summaries through graphical and textual methods.
We introduce a data-driven, model-agnostic technique for generating a human-interpretable summary of the salient points of contrast within an evolving dynamical system, such as the learning process of a control agent. It involves the aggregation of transition data along both spatial and temporal dimensions according to an information-theoretic divergence measure. A practical algorithm is outlined for continuous state spaces, and deployed to summarise the learning histories of deep reinforcement learning agents with the aid of graphical and textual communication methods. We expect our method to be complementary to existing techniques in the realm of agent interpretability.