Derivative-Free Optimization via Proximal Point Methods

Abstract

Derivative-Free Optimization (DFO) examines the challenge of
minimizing (or maximizing) a function without explicit use of derivative
information. Many standard technique in DFO are based on using model
functions to approximate the objective function, and then applying
classic optimization method on the model function. Interestingly,
research into DFO methods has not considered how the classical proximal
point method might be adapted to a derivative-free setting. This is
surprising, as the proximal point method provides an elegant, flexible,
and robust framework, both in theory and practice. In this talk we
examine the potential to apply Proximal Point methods in Derivative Free
Optimization.

Based on joint work with Yves Lucet.

Details