![]() Incorporating bound (lb and ub) and linear constraints (A, Aeq, b, beq) into my YALMIP problem formulation is simple. I would like to be able to use any existing nonlcon function which is already directly suitable for use in FMINCON (or KNITRO, which is the same). I would like to add other solvers, such as PENLAB, in order to support LMIs and BMIs in addition to nonlcon style nonlinear constraints. I call those solvers directly without using YALMIP. I just pass though the nonlcon function handle (with adjustment to the bounds if calling BARON in order to match its nonlinear constraint conventions). I have a special purpose optimizer still under development for which I currently have FMINCON, KNITRO, and BARON as solvers available for subproblems having nonlinear constraints. No, this is my own question which I've had on the back burner for a while. Assume that all the code in the nonlcon function to create a c (inequality with rhs = 0) or ceq (equality to 0) would be valid YALMIP code if entered in the normal way in YALMIP.Ĭonstraints = įunction = nonlcon_example(x,p)Ĭ = I would like to add a capability to allow solvers to be called via YALMIP, making use of the nonlcon function handle. ![]() The function handle of the nonlinear constraints function is an input to my MATLAB function, and is passed though to any solver called directly, and that works fine. My paradigm is to use this within my own MATLAB function which, in order o solve subproblems, either calls a solver directly (i.e., without using YALMIP) or via YALMIP. There may or may nor be gradients provided as optional 3rd and 4th arguments, which I suppose YALMIP would be free to ignore in favor of its own gradient computation. Suppose I have nonlinear constraints specified in a nonlcon constraints function in FMINCON form.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |