VXQR1 performs an approximate unconstrained minimization of a not necessarily smooth function of many continuous arguments. No gradients are needed. A limited amount of noise is tolerated.

The primary goal is to get in high dimensions quick improvements with few function evaluations, not necessarily to find the true global minimum.

A bounded initial search region must be specified but the actual search is not confined to this region; thus the best point found may be outside the search region.

The program was written by Arnold Neumaier (University of Vienna). Please inform the author at Arnold.Neumaier@univie.ac.at if you make serious use of this code.

The gzipped tar file v1.0.tar.gz (16K) contains the Matlab source code for Version 1.0 (2010), a driver program showing its use, and the paper

The individual files in the package can be viewed here.

This paper describes the method implemented and some test results. Please quote this paper when using this package in scientific work.

Progress figures for each of the 25 runs on the collection of test problems from Scalability of Evolutionary Algorithms and other Metaheuristics for Large Scale Continuous Optimization Problems, ( in dimensions 50, 100, 2000, 500, and 1000)

Cumulative result tables for these runs, and a comparison with DE, CHC, and CMA-ES.

Additional statistics for these comparisons

Global (and Local) Optimization
Statistics Links
Mathematical Software

my home page (http://arnold-neumaier.at)

Arnold Neumaier (Arnold.Neumaier@univie.ac.at)