Generative Models to Guide Vision Studies

The probabilistic brain hypothesis posits that perception and behavior emerge from an internal world model that allows to perform inference of its state and enable predictions. In this talk, I will introduce the concept of generative models and their fundamental role in vision science. By framing perception as an inference process, generative models provide a principled approach to understanding how the brain handle representations from ambiguous sensory input. I will discuss why having interpretable generative models is crucial for advancing our understanding of vision and behavior, highlighting key limitations and how to overcome those.