How to use patternsearch to optimize two control vectors of a function; optimizing both variables instead of passing one of them?
Show older comments
Dear Alan, I have an optimization problem where my function has two control variables as the following
function [NetPV,finalSupply] = OptimizedRate (Weights,NRates)
Weights and NRates are vectors as: Weights = [x1 x2 x3] and NRates = [y1 y2 y3]
I want to optimized Weights and NRates not just passing one of them. I have seen some examples like: https://uk.mathworks.com/help/optim/ug/passing-extra-parameters.html
and tried them. However this pass NRates variable instead of optimizing it. [x,fval,exitflag,output] = patternsearch(@(Weights)OptimizedRate(Weights,NRates),x0,[],[],Aeq,beq,LB,UB,options);
I'm looking for away to optimize Weights and NRates variables jointly.
Thanks for your help!
Accepted Answer
More Answers (1)
Alan Weiss
on 4 Jan 2018
I think that the simplest way is to combine your variables into one vector, x = [Weights,NRates], which is what any optimization solver needs. See Compute Objective Function.
For example,
function [NetPV,finalSupply] = mynewfun(x)
Weights = x(1:3);
NRates = x(4:6);
[NetPV,finalSupply] = OptimizedRate(Weights,NRates);
Optimize mynewfun instead of OptimizedRate. Be sure to give your initial point x0 as a 6-element vector.
Alan Weiss
MATLAB mathematical toolbox documentation
3 Comments
Yaser Khojah
on 4 Jan 2018
Alan Weiss
on 4 Jan 2018
The linear constraints should be written in terms of the 6-D variable x.
Alan Weiss
MATLAB mathematical toolbox documentation
Yaser Khojah
on 4 Jan 2018
Categories
Find more on Get Started with Optimization Toolbox in Help Center and File Exchange
Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!