Free Energy Principle as a Universal Theory of Brain Function
The free energy principle proposes that all cognitive and perceptual systems, from simple organisms to human brains, function as prediction machines that continuously minimize a quantity called variational free energy.
Generative and Recognition Models in the Brain
Populations of latent neurons within cortical hierarchies act as hidden variables that summarize the causes of sensory data, while distinct neural circuits implement generative and recognition mappings between latent and sensory levels.
Priors, Beliefs, and Context in Predictive Brains
Brains accumulate prior beliefs about the world—statistical regularities distilled from evolution and experience—and use them to interpret ambiguous sensory inputs in context-dependent ways.
Perception as Inference Under the Free Energy Principle
Latent neurons encoding candidate causes and sensory neurons conveying raw data engage in a continual inferential loop, adjusting activity to find explanations that best reconcile expectations with observations.
Hollow Mask Illusion and Free Energy Minimization
The hollow mask illusion, where a concave mask of a face appears convex when rotated, reveals how strongly high-level priors about faces constrain visual perception.