Self-Attentional Credit Assignment for Transfer in Reinforcement Learning
This work addresses the challenge of knowledge transfer for general learning agents in reinforcement learning, presenting a new method that is incremental in its approach to enhancing existing algorithms.
The paper tackles the problem of sample inefficiency in reinforcement learning by proposing a novel transfer learning approach that uses backward-view credit assignment to identify structural invariants across tasks, resulting in improved sample efficiency without specifying concrete numbers.
The ability to transfer knowledge to novel environments and tasks is a sensible desiderata for general learning agents. Despite the apparent promises, transfer in RL is still an open and little exploited research area. In this paper, we take a brand-new perspective about transfer: we suggest that the ability to assign credit unveils structural invariants in the tasks that can be transferred to make RL more sample-efficient. Our main contribution is SECRET, a novel approach to transfer learning for RL that uses a backward-view credit assignment mechanism based on a self-attentive architecture. Two aspects are key to its generality: it learns to assign credit as a separate offline supervised process and exclusively modifies the reward function. Consequently, it can be supplemented by transfer methods that do not modify the reward function and it can be plugged on top of any RL algorithm.