Motivation:

1. To implement the paper of Shakir Mohamed and Danilo Rezende https://arxiv.org/abs/1509.08731 aka 'Variational Information Maximisation for Intrinsically Motivated Reinforcement Learning'
2. To develop an understanding of the value and limitations of Empowerment for Intrinsic Control.
3. Empowerment was originally introduced by Klyubin, Polani and Nehaniv (2005) here: https://ieeexplore.ieee.org/document/1554676. A more comprehensive treatment from Salge, Glackin and Polani (2014) here: https://arxiv.org/abs/1310.1863

Note: This project was initially co-developed with Ildefons Magrans([email protected]).