Bellman documentation

Bellman is a Python toolbox for model-based reinforcement learning (MBRL). That is, for solving sequential decision problems by learning predictive models of parts of the problem.

The toolbox extends the open source TensorFlow-Agents library, initially conceived to tackle model-free RL problems, to enable model-based RL capabilities.

Getting started

Start with the Approximating MDPs Jupyter Notebook for an overview of the predictive models.

Alternatively, read the API reference documentation.

bellman

API reference documentation for the Bellman toolbox.