What is the benefit to supply the gradient for function minimization?

Question

Pasquale 2017-11-7

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/365660-what-is-the-benefit-to-supply-the-gradient-for-function-minimization

编辑： Honglei Chen 2017-11-7

Do you know if is it better to supply the gradient using 'fminunc'? For better I mean the time spent for analysis, and its accuracy.

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

Honglei Chen 2017-11-7

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/365660-what-is-the-benefit-to-supply-the-gradient-for-function-minimization#answer_289895

编辑：Honglei Chen 2017-11-7

I'm not an expert on this, but I believe that if you don't provide the gradient function, then it has to be numerically estimated from your function. Thus providing the gradient (if available) can help the algorithm to converge faster.

1 个评论
显示 -1更早的评论隐藏 -1更早的评论

Pasquale 2017-11-7

I can valuate and provide the gradient, but it comes out from a lot of calculations, and my whole code runs on for loops with more than 1000 elements. So I don't know if it is more expensive to calculate the gradient through code lines, or let the solver fminunc do this. I should try and compare

请先登录，再进行评论。