Adam Koblinger

Ádám Koblinger is interested in understanding the computational mechanisms that allow humans and animals to efficiently process inherently ambiguous and noisy sensory information, and how these computations are implemented in the brain. He uses probabilistic Bayesian models to address these questions at multiple levels of analysis. At the computational level, he develops models describing how humans integrate uncertain information from sequential observations in environments with unstable statistical properties. At the algorithmic level, he investigates efficient algorithmic realizations of abstract Bayesian computations that can account for various top-down effects observed in the visual cortex. More specifically, he tests task-dependent sampling strategies optimized to represent complex posterior distributions in resource-constrained inference systems.