Large-Scale Nonlinear Programming

Northwestern University, Evanston IL

Investigators

Abstract

Large-Scale Nonlinear Programming Large nonlinear optimization problems arise in many important areas of application such as climate modeling, weather forecasting, network design, and finance. An important open problem is how to solve these problems reliably and quickly. The proposal describes a variety of new algorithms, based on barrier techniques, that promise to be effective on problems of unprecedented size and complexity. To ensure that the new methods are robust even in the case of high ill-conditioning, novel regularization techniques are proposed. An Internet-based solver implementing the new algorithms will be developed as part of this project. It will allow the engineering and scientific communities to solve large optimization problems automatically using a web-based interface.