Manfred Opper

Statistical Physics, Information Theory, applied Probability and their application to machine learning, disordered materials and other compex systems. Current projects include: * Analysis of the generalization ability of neural nets and other learning machines using methods of Statistical Physics. * General Bounds on entropic error measures in estimating probability distributions. * Worst Case over sequence prediction. * Mean Field methods in probabilistic modelling. * Bayesian approaches to online learning. * Nonequilibrium dynamics of disordered systems. * Support Vector Machines * Population dynamics.