Combining Behaviors with the Successor Features Keyboard
This work addresses the cumbersome need for hand-designed representations in transfer learning for reinforcement learning, offering an incremental improvement by automating feature discovery.
The paper tackled the problem of transferring behavioral knowledge across tasks in reinforcement learning by proposing the Successor Features Keyboard (SFK) with the Categorical Successor Feature Approximator (CSFA), enabling transfer with discovered state-features and task encodings in a challenging 3D environment, where SFK transferred most quickly to long-horizon tasks compared to baselines.
The Option Keyboard (OK) was recently proposed as a method for transferring behavioral knowledge across tasks. OK transfers knowledge by adaptively combining subsets of known behaviors using Successor Features (SFs) and Generalized Policy Improvement (GPI). However, it relies on hand-designed state-features and task encodings which are cumbersome to design for every new environment. In this work, we propose the "Successor Features Keyboard" (SFK), which enables transfer with discovered state-features and task encodings. To enable discovery, we propose the "Categorical Successor Feature Approximator" (CSFA), a novel learning algorithm for estimating SFs while jointly discovering state-features and task encodings. With SFK and CSFA, we achieve the first demonstration of transfer with SFs in a challenging 3D environment where all the necessary representations are discovered. We first compare CSFA against other methods for approximating SFs and show that only CSFA discovers representations compatible with SF&GPI at this scale. We then compare SFK against transfer learning baselines and show that it transfers most quickly to long-horizon tasks.