Nonlinear Distributional Gradient TemporalDifference Learning
[edit]
Proceedings of the 36th International Conference on Machine Learning, PMLR 97:52515260, 2019.
Abstract
We devise a distributional variant of gradient temporaldifference (TD) learning. Distributional reinforcement learning has been demonstrated to outperform the regular one in the recent study \citep{bellemare2017distributional}. In the policy evaluation setting, we design two new algorithms called distributional GTD2 and distributional TDC using the Cram{é}r distance on the distributional version of the Bellman error objective function, which inherits advantages of both the nonlinear gradient TD algorithms and the distributional RL approach. In the control setting, we propose the distributional GreedyGQ using similar derivation. We prove the asymptotic almostsure convergence of distributional GTD2 and TDC to a local optimal solution for general smooth function approximators, which includes neural networks that have been widely used in recent study to solve the reallife RL problems. In each step, the computational complexity of above three algorithms is linear w.r.t. the number of the parameters of the function approximator, thus can be implemented efficiently for neural networks.
Related Material


