This project aims at automating the process of analysis and design of optimization algorithms. The basis of the project is to formulate the problem of analyzing convergence of an optimization algorithm as another optimization problem. This problem finds the worst case function, within the class of considered functions, given a performance metric. Based on this, we develop methodologies for finding Lyapunov functions for proving convergence, devise new algorithms, and find extended convergent parameter regions for existing methods.
Gällande start-/slutdatum2020/08/20 → …