Friston’s Free Energy Principle Explained (part 1)

unread

https://jaredtumiel.github.io/blog/2020/08/08/free-energy1.html

By Jared Tumiel. A friendly but rigorous guide to Karl Friston’s Free Energy Principle (FEP).

Covers:

  • Bayesian inference
  • ’Phenotype’ as a set of viable states
  • Entropy and expected surprise
  • Kullback-Leibler Divergence
  • Recognition- and Generative-densities
  • Derivation of the ‘free energy’ term

Goal of the series: Build toward implementing Active Inference under FEP in Python. Draws on statistical mechanics, reinforcement learning, neuroscience, predictive coding, information theory, variational inference, and embodied cognition.