- subjected to
- (continuous functions,
- sometimes with some numerical noise)
In NSP default stencil value (for finite-difference derivatives approximation, if derivatives are not supplied by user or FuncDesigner automatic differentiation) is 3, while in NLP it is 1 (see DerApproximator documentation). Also, word NSP instead of NLP informs other programmers (who could edit your code) that your problem is nonsmooth.
|ralg||BSD||Dmitrey||For medium-scaled problems (nVars up to ~1000...1500); ill-conditioned, piecewise linear and polynomial, non-smooth & noisy ones are also allowed. All types of constraints (lb <= x <= ub, A*x <= b, Aeq*x = beq, c(x) <= 0, h(x) = 0). r-algorithm with adaptive space dilation had been invented by Ukrainian academician Naum Z. Shor (my teacher //Dmitrey). This solver is enhanced from time to time (almost each OpenOpt release).|
|interalg||BSD||Dmitrey||Solver with guaranteed user-defined accuracy fTol: abs(f - f*) < fTol. Handling of general constraints is available since r. 0.36|
|gsubg||BSD||Dmitrey||For large-scaled convex problems with guaranteed user-defined precision. Lots of work still remains to be done, especially for constrained problems.|
Currently unconstrained only, for medium-scaled problems (nVars up to ~1000...1500)
Requires known fOpt (optimal value) and fTol (required objective function tolerance, may be very small).
|ShorEllipsoid||BSD||Dmitrey||currently it's a tentative implementation of Naum Z. Shor method of ellipsoids; it's unconstrained, for small-scale problems with nVars = 1..10, requires r0: norm(x0-x*)<r0)||No|
|scipy_fmin||BSD||an implementation of Nelder-Mead simplex algorithm; unconstrained, cannot handle user-supplied derivatives|
|sbplx||LGPL||Steven G. Johnson.||A variant of Nelder-Mead algorithm. Requires nlopt installed. Can handle box bound constraints.|