Active Inference — Variational Bayes Perception

An agent minimizes variational free energy F = KL[q(s)‖p(s|o)] + log p(o) to infer hidden states s from noisy observations o. Watch the posterior belief (green) converge to the true state (white dot) under different precision (noise) conditions.

F = — | Belief: μ=(—,—) | Error: —