About the event
In this talk, Alexandre d'Aspremont will describe a convergence acceleration technique for generic optimisation problems. This scheme computes estimates of the optimum from a nonlinear average of the iterates produced by any optimisation method. The weights in this average are computed via a simple linear system, whose solution can be updated online. This acceleration scheme runs in parallel to the base algorithm, providing improved estimates of the solution on the fly, while the original optimisation method is running. Numerical experiments are detailed on classical classification problems. Biography Alexandre d’Aspremont is a CNRS researcher in the Département d’Informatique at Ecole Normale Supérieure, Paris, France. He obtained his PhD from Stanford University, and subsequently held Assistant and Associate Professorships at Princeton University. Alexandre’s research interests broadly span optimisation and statistical learning. This talk is aimed at anyone with an interest in optimisation.