Web: http://arxiv.org/abs/2201.07296

Jan. 24, 2022, 2:10 a.m. | Bekzhan Kerimkulov, James-Michael Leahy, David Šiška, Lukasz Szpruch

stat.ML updates on arXiv.org arxiv.org

We study the global convergence of policy gradient for infinite-horizon,
continuous state and action space, entropy-regularized Markov decision
processes (MDPs). We consider a softmax policy with (one-hidden layer) neural
network approximation in a mean-field regime. Additional entropic
regularization in the associated mean-field probability measure is added, and
the corresponding gradient flow is studied in the 2-Wasserstein metric. We show
that the objective function is increasing along the gradient flow. Further, we
prove that if the regularization in terms of the …

arxiv cross entropy gradient math network neural neural network policy

More from arxiv.org / stat.ML updates on arXiv.org

Engineering Manager, Machine Learning (Credit Engineering)

@ Affirm | Remote Poland

Sr Data Engineer

@ Rappi | [CO] Bogotá

Senior Analytics Engineer

@ GetGround | Porto

Senior Staff Software Engineer, Data Engineering

@ Galileo, Inc. | New York City or Remote

Data Engineer

@ Atlassian | Bengaluru, India

Data Engineer | Hybrid (Pune)

@ Velotio | Pune, Maharashtra, India