Distributional Multivariate Policy Evaluation and Exploration with the Bellman GAN

The dataset is used to evaluate the distributional approach to reinforcement learning (DiRL) and its equivalence to Generative Adversarial Networks (GANs).

BibTex: