Global Optimization Test Results

Global Search, As Timely As Ever

``Consider everything. Keep the good. Avoid evil whenever you notice it.''
(1 Thess. 5:21-22)

This file is part of my global optimization web site.

Please submit test results (latex source, ASCII, or HTML format) for your own algorithms on any of the test sets in the Global Optimization Test Set Collection. Test results will be made available here.

The idea is that you can enter a competition on the fastest algorithms and you can look up the results others obtained with their favorite methods. You can also submit your disappointments with publicly available codes; this will give a challenge to developers of better methods.

I'd like to encourage a common format for reporting results and invite your ideas about what you would like to see in a result database. To get you started, look at my current ideal for Test Result Presentation

Comparison of derivative-free optimization algorithms (2012, by Nick Sahinidis)

Black-Box Optimization Benchmarking (BBOB) 2012 (by Auger, Hansen, et al.)

Comparison of Global Solvers
Within the COCONUT project, we performed extensive tests of the global solvers BARON, GlobSol, ICOS, LGO, LINGO, OQNLP, Premium Solver, MINOS, and COCOS (the solver in the alpha test version of the COCONUT environment) for various sublibraries of the COCONUT benchmark.

A Comparison of Stochastic Global Optimization Programs in Fortran or C on 45 bound-constrained problems of dimension 2-30 (by Erich Janka)
Of the programs tested, the Derivative-Free Boender-Timmer-Rinnoy Kan Algorithm by Tibor Csendes is the clear winner. (But MCS and DIRECT/gblSolve that are publicly available in Matlab only and did not fit the environment of the above study are even better, at least on the more limited Dixon-Szego test set.)

W. Huyer, A comparison of some algorithms for bound constrained global optimization, Manuscript (2004). (pdf file, 69K)
The paper contains comparisons of the three heuristic global optimization programs MCS, Global, PGSL, using black box function evaluation only.

Comparison of results on the 2006 CEC Benchmark function set for the
2006 CEC Benchmark for constrained optimization

Cambridge Cluster Database (The Lennard-Jones cluster problem and more realistic cluster problems, with record lists)

Cluster with minimum LJ potential (record lists)

The Ph.D. thesis of Tom Epperly contains problem definitions and solutions (ps.gz, 59K) for constrained global optimization problems (among others from the Pardalos-Rosen test collection) together with test results for his branch and bound algorithm.

Some Branch and Bound Results (reported by Baker Kearfott)

Some Branch and Bound Results (reported by Peter Spellucci)

Some Simulated Annealing Results (reported by Philippe Bonnardel)

Some Results by George Bilchev obtained with his prize winning inductive search algorithm, for my Simple Challenges for Global Optimization Codes

Some Genetic Algorithms Results for simple problems up to dimension 1000
(my summary of a collection by Leo Lazauskas - unfortunately no longer available)

Some results for RGA (Regularized Global Approximation, reported by Gennady Ryzhikov)

Test Results for Local Optimization Codes

Benchmarks for Optimization Software (collected by Mittelmann and Spellucci)

Test Results for Moré/Garbow/Hillstrom test set and some problems from the CUTE test set (collected by Alexei Kuntsevich)
local copy (of old version; obsolete)

Testing L-BFGS-B versus MINOS on the bound constrained problems from CUTE

Some results on semidefinite programming codes (.ps, 194K)