How to get the step size by steepest-descent method in Matlab?(Updating the step size simultaneously)

1 次查看（过去 30 天）

Jinghua Li 2016-11-23

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/313641-how-to-get-the-step-size-by-steepest-descent-method-in-matlab-updating-the-step-size-simultaneously

关闭： Walter Roberson 2016-11-24

I want to optimization parameters of Gaussian membership function with steepest-descent method in iterative process. Unfortunately, I can’t define the parameters of step size because of the changeable number of the parameter with different data set. How to get the step size of steepest-descent method by one-dimensional linear search is a current problem for me. What’s more, how to update the step size simultaneously if the object function consists of N Gaussian function? (N is a variable and decided by your data set.) Looking forward to your guidance!

f=exp(-(x1-mul1)^2/(2*sigma1^2))*exp(-(x2-mul2)^2/(2*sigma2^2))*…*exp(-(xN-mulN)^2/(2*sigmaN^2)); Where X(X=[x1,x2,…,xN]) is obtained from input, so we can think X is a known data set. Mul and sigma are parameters, which are objections improved by steepest-descent method.