Structural Credit Assignment in Neural Networks using Reinforcement Learning

no code implementations • NeurIPS 2021 • Dhawal Gupta, Gabor Mihucz, Matthew Schlegel, James Kostas, Philip S. Thomas, Martha White

In this work, we revisit this approach and investigate if we can leverage other reinforcement learning approaches to improve learning.

reinforcement-learning Reinforcement Learning (RL)

Classical Policy Gradient: Preserving Bellman's Principle of Optimality

no code implementations • 6 Jun 2019 • Philip S. Thomas, Scott M. Jordan, Yash Chandak, Chris Nota, James Kostas

We propose a new objective function for finite-horizon episodic Markov decision processes that better captures Bellman's principle of optimality, and provide an expression for the gradient of the objective.

Learning Action Representations for Reinforcement Learning

no code implementations • 1 Feb 2019 • Yash Chandak, Georgios Theocharous, James Kostas, Scott Jordan, Philip S. Thomas

Most model-free reinforcement learning methods leverage state representations (embeddings) for generalization, but either ignore structure in the space of actions or assume the structure is provided a priori.

reinforcement-learning Reinforcement Learning (RL)