More Jorge J., Wright Stephen J. Optimization Software Guide

Society for Industrial and Applied Mathematics, 1993. — 167 p.This book presents information on the current state of numerical optimization software that we collected in preparation for a short course on Numerical Optimization Algorithms and Software. This information was presented at the SIAM Optimization meeting in May 1992 and at the SIAM Annual meeting in July 1992.
Part I contains an overview of algorithms for different classes of optimization problems. The treatment here is by no means exhaustive. We aim instead to give the reader a feeling for each of the algorithms implemented in the software of Part II, along with pointers to books and survey papers in which more thorough and rigorous discussions can be found. We hope that Part I includes enough information to allow the reader to identify software packages worthy of further investigation.
Part II comprises product descriptions of about a page each, provided by software vendors and individual researchers. Although we made an effort to contact as wide and diverse a collection of vendors as possible, our survey of optimization software is by no means complete. We hope that we have listed the most widely used packages within each category, along with some newer or less well known entries. We plan to expand and update the listings as further information becomes available.
We have refrained from making comparisons of the software on the basis of performance. This task is clearly impossible for the large number of packages included in this guide. Even if a standard test set were available—which is not the case for most problem classes—the continuing evolution of most software packages would make any comparison obsolete within months.
We have also omitted information on pricing of the software. Prices tend to change even more rapidly than the software itself and vary according to whether the buyer is in academia or industry.
This book does not mention software for such widely studied classes of problem as stochastic programming, optimal control, and parameter estimation. This is because no simple, catch-all mathematical formulation exists for each of these classes that can serve as a basis for efficient algorithm design. For example, to devise an efficient algorithm for an optimal control problem, we must know whether the dynamics of the underlying system are governed by ordinary or partial differential equations, and whether these dynamics are stochastic or deterministic. In stochastic programming, the choice of algorithm may differ according to whether the problem is two-stage or multistage, and according to whether the cost functions are linear or nonlinear. Certainly, problems such as these can be formulated as general optimization problems,
but the price paid for ignoring the special structure is a significant (often ruinous) loss of efficiency