Katya Scheinberg: Introduction to derivative-free and zeroth order optimization I скачать в хорошем качестве

Katya Scheinberg: Introduction to derivative-free and zeroth order optimization I

Katya Scheinberg Aisenstadt Chair Lectures(20-30 Mai 2025/ May 20-30 2025) Série de conférences de la Chaire Aisenstadt https://www.crmath.ca/en/activities/#... Katya Scheinberg (Georgia Institute of Technology) May 21, 2025 Title : Introduction to derivative-free and zeroth order optimization Abstract: We will introduce the derivative free optimization which is an area of optimization that studies optimization methods that do not rely on access to differentiation of the objective. In the first lecture we will discuss methods that are truly zeroth-order - that is methods that do not attempt to approximate derivative information. These methods will include Nelder-Mead method, direct search methods and patterns search methods. We will discuss their convergence and complexity. In the second lecture we will overview methods that build models of the objective function and thus approximate its first and, sometimes, second derivatives.