Optimization Technique

Hi, I am using a formula for response Y dependent on variables R, T, h, and L which goes like this Y = Const + exp( C1*(R-C2)^2+ C3*(T-C4)^2+ C5*(h-C6)^2 C7*(L-C8)^2).... ( many exponential terms follow) I need to know what are the optimization techniques available with MAT LAB to find optimum value of Y given that I know the constrains on R, T, h, and L?

Answer 1

[Seth DeLand]

If you're trying to minimize or maximize Y, and Y is a scalar, I'd recommend trying FMINCON from the optimization toolbox first. This is a gradient-based method, so it will travel downhill to find the local minimum value of Y. You can constrain the variables with bounds, linear equalities/inequalities, and nonlinear equalities/inequalities.

If Y is a vector, and you're interested in a least square minimization of Y, then the LSQNONLIN solver is probably more along the lines of what you're looking for.

Depending on what your function for Y looks like, you might also want to try some of the solvers from the global optimization toolbox. These solvers were designed to search for global minimums of nonlinear problems. I recommend starting with the PATTERNSEARCH solver if you're interested in using global optimization techniques.

[Nihal]

Thanks Seth. Could you also tell me that what functions could I use if I don't want to minimize Y but to match a target?