Mathematical Modeling of Perceptions and Emotions

(An Approach using Information Theory)

Abstract

Information-theoretic free-energy means the difference between recognition model and generative model. The principle of free-energy minimization suggests that the recognition model must follow a Bayesian posterior. I consider that the minimized free-energy represents potential information contents to be processed in the brain after recognizing external physical phenomena. I discuss how the minimized free-energy predicts human emotions such as surprise and valence (i.e. positivity and negativity). The minimized free-energy can be decomposed into two terms: novelty and perceived complexity. I demonstrate that the summation of the two terms works as arousal potential and forms a valence function.