sschwoebel / activeinferencebpbethe

Python implementation of the Paper "Active inference, belief propagation, and the Bethe approcimation"

ActiveInference, Belief Propagation, and the Bethe Approximation

This repository implements the active inference update equations as derived in (Schwöbel et al. 2018). It provides several classes which calculate beliefs of an artificial agent in accordance with free energy minimization. Additionally, the file 'run_example_gridworld.py' sets up different agents for the conditions described in the paper.

You are free to use and modify the code. If you use it for another publication, please cite our article.

If you have any questions regarding the code, the paper, or the equations, please do not hesitste to contact me (@SSchwoebel).

Schwöbel, S., Kiebel, S., & Marković, D. (2018). Active inference, belief propagation, and the bethe approximation. Neural computation, 30(9), 2530-2567.