How to resume genetic algorithm after x number of generations?

Question

Joe 2016-2-23

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/269669-how-to-resume-genetic-algorithm-after-x-number-of-generations

回答： Alan Weiss 2016-2-24

Hi - I want to do a generic algorithm optimization in batches of generations (e.g run 10 generations, then resume and do other 10, etc).

I am using the regular command line, getting the final population and final scores after each run, and then I inject those as initial population and initial score of the next run. this mostly solves my need.

However, I don't know how to handle the initial penalty value. I can't find in the output structure any final penalty value...

I would like to introduce add the initial penalty value of the next run the final penalty value of the previous one.

Any suggestions on how to make it happen?

0 个评论
显示 -2更早的评论隐藏 -2更早的评论

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Alan Weiss 2016-2-24

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/269669-how-to-resume-genetic-algorithm-after-x-number-of-generations#answer_211047

The penalty value relates only to nonlinear constraints. You are correct, the penalty value is not reported in the output structure, so you won't be able to handle a restart exactly.

However, if you are using a nonlinearly constrained problem, I am very surprised that you can even run for 10 generations. Usually, the nonlinear constraint algorithm leads to a solution with just a few generations, maybe 5, but each generation does quite a few subproblem solutions involving a lot of generations. So I am not sure what you are doing, because I think that the output function reports after major iterations, and there would only be 5 or so of those.

Alan Weiss

MATLAB mathematical toolbox documentation