Our group explores theoretical principles of neural computations with a particular focus on sensory systems. We study how complex natural stimuli such as sounds or images are represented in neural activity, how does the brain use such representations to infer the state of the environment, and how do such inferences support flexible behavior. We rely on information theory, computer science and probabilistic machine learning to develop optimal information processing strategies which might be approximated by biological systems. We then confront theories with reality in close collaborations with experimental groups. Our hope is that this approach will bring us towards identifying general rules governing information processing in living organisms.