Machine Learning NeEDS Mathematical Optimization with Prof Fabio Schoen

Abstract: In this talk I will start by showing how it is possible to use clustering methods to improve some Global Optimization (GO) methods; this idea was first proposed in the very beginning of GO in the 70’s, but later it was abandoned. We re-discovered this methodology in recent years and produced variants which, for some problem classes, are quite promising. One of such classes is that of detecting the minimum energy configuration of clusters of atoms.

In the second part of the talk, I will consider clustering as a hard to solve GO optimization problem itself and will outline some ideas on how to modify some standard GO heuristics in order to make clustering more efficient.