Hi all, I would like to perform some benchmarks over a set of nonlinear problems where some functions (objective or nonlinear constrains) are defined over restricted domains, e.g. an_objective  > min wrt x>0 x + 5y + 0.01 * x * sqrt(x) < 10 I would easily write my own set of tests (there are some already done), but since my solver is involved (http://openopt.org/ralg, latest stable release vs my current implementation vs some other solvers) it would be much better to have a set taken from an external source, preferably an article or a book. I have searched google with "finite domain non linear benchmark" but there are no appropriate results (I don't know about some nonfree journals, however). So, does anyone know any appropriate source? Thank you in advance, D. asked 10 Apr '10, 09:43 Dmitrey 
Assuming that by "finite domain" you mean bounded domain (as opposed to solving over a finite lattice), there are several libraries of NLP test problems linked off the NEOS wiki. Many (most?) of them are constrained, including nonlinear constraints (and, I think, some with nonconvex feasible regions). answered 10 Apr '10, 14:29 Paul Rubin ♦♦ Sorry, I meant bounded domain with some constraints defined over restricted domain, e.g. sqrt or log are not defined over R. Unfortunately, I'm not familiar with the data formats there, and, as far as I understood, most of them are defined over whole R^nVariables. Thus I'm still searching for text description of the test cases with restricted domains, if anyone knows.
(10 Apr '10, 15:17)
Dmitrey
