fmincon and fmiunc gives different result for the same problem

5 views (last 30 days)
Can anyone help how to setup the result from fmincon and fminunc to give the same result for same problem ?
Thanks

Accepted Answer

Walter Roberson
Walter Roberson on 29 Nov 2020
This is expected. fmincon() and fminunc() use different algorithms.
You might possibly be able to get somewhere by forcing fminunc to use 'trust-region' algorithm, which requires that your function outputs the gradient as its second output, and requires some flags set in the options. You would in turn configure fmincon to use 'trust-region-reflective'; again that requires that your function outputs the gradient as its second output, and requires some flags set in the options.
It is not entirely clear to me that the 'trust-region' algorithm of fminunc is the same as 'trust-region-reflective' of fmincon .
  1 Comment
HN
HN on 29 Nov 2020
Thank you so much Walter Roberson ,
I also have two more quesions.
  1. Fminunc option I used can only dispaly the final optimum point or cannot list all searched points toward the optimum point. How can I handle that. or
  2. Can I automatically pick the minimum value of the funciton and corresponding optimum point ?
Thank you

Sign in to comment.

More Answers (0)

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!