Michael's Home Page: Nonlinear Programming Research


My interest in nonlinear programming algorithms grew out of my research in transportation analysis. A common thread in my research endeavours is a wish to not only contribute to scientific knowledge, but also (and a wish that is equally strong) to place the existing knowledge in a unified framework; this is, in my view, a far too overlooked part of scientific research. My research has so far led to the formulation and analysis of a general class of descent algorithms for nonlinear programming problems, the class of cost approximation algorithms. This class includes many classical algorithms, such as steepest descent and Newton's algorithm in unconstrained optimization, the Frank-Wolfe and gradient projection algorithms in differentiable, constrained optimization, and regularization algorithms such as the proximal point algorithm for general nondifferentiable optimization. The algorithm class has also been extended to the solution of variational inequalities. A second, and much more recent, interest, is in applications of mathematical programming algorithms. Here you will find both reports related to the class of cost approximation algorithms (including my PhD thesis) and recent applications oriented ones. The latest addition (2005) is a report on a core problem, namely one that often is referred to as the (continuous) nonlinear knapsack problem. "Not much new is under the sun" is a rather good summary of this work.

Note: the preliminary reports may not be available yet. Be patient; one of these days they will be ...


Text Books:

Monograph:

Dissertation:

Reports on Cost Approximation Concepts:

Reports on Nondifferentiable Optimization:

Reports on Sensitivity Analysis:

Reports on Stochastic Bilevel Programming:

Reports on Engineering Applications:

Reports on Column Generation Algorithms:

Reports on Resource Allocation Problems:

Reports on Nonlinear Network Flow Problems: