Orchestrated Value Mapping for Reinforcement LearningDownload PDF


Sep 29, 2021 (edited Oct 05, 2021)ICLR 2022 Conference Blind SubmissionReaders: Everyone
  • Keywords: Reinforcement Learning, Value Mapping, Reward Decomposition
  • Abstract: We present a general convergent class of reinforcement learning algorithms that is founded on two distinct principles: (1) mapping the value function into a different space via arbitrary functions from a broad class and (2) linearly decomposing the reward signal into multiple channels. The first principle enables asserting specific properties on the value function that can enhance learning. The second principle, on the other hand, allows for the value function to be represented as a composition of multiple utility functions. This can be leveraged for various purposes, including dealing with highly varying reward scales, incorporating a priori knowledge about the sources of reward, and ensemble learning. Combining the two principles yields a general blueprint for instantiating convergent algorithms by orchestrating diverse mapping functions over multiple reward channels. This blueprint generalizes and subsumes algorithms such as classical Q-learning, Logarithmic Q-learning, and Q-Decomposition. Moreover, our convergence proof for this general class relaxes certain required assumptions in some existing algorithms. Using our theory we discuss several interesting configurations as special cases. Finally, to illustrate the potential of the design space that our theory opens up, we instantiate a particular algorithm and evaluate its performance on the suite of Atari 2600 games.
  • One-sentence Summary: We present a general convergent class of RL algorithms based on combining arbitrary value mappings and reward decomposition.
0 Replies