Adaptive coding for dynamic Bayesian inference

Organisms must infer the properties of their ever-changing environment from sensory stimuli. Representing these stimuli is complicated, costly and requires metabolic resources. In this talk I will discuss how neurons and other biological systems might minimize the effort to accurately represent sensory signals without losing their ability to keep the track of the changing environment. Our apporach blends dynamic Bayesian inference, information theory and a lot of speculation.