- #1
noon0788
- 22
- 0
I have a user-defined function that doesn't seem to work with NMinimum, FindMinimum, or any of the other optimization commands. I think this is due to the fact that the function uses elliptic integrals and matricies to obtain an output. It has multiple variables, but I want to keep them all constant and minimize one. Does anyone know how to minimize ANY function? We'll just treat it as a black box, receiving inputs and producing outputs.
Oh, and also, the domain of the function is constricted. I don't know how to calculate the domain of the function either.
I've tried using a gradient search method already. I find the gradient at a point by calculating two points and doing it the old-fashioned way. Then I subtract the gradient from its point to get a new point. I keep searching this way until the gradient is relatively small (.000001). However, the code breaks a good deal because when the gradient is extremely large, this sends the calculated, new point out of the function's domain, and everything goes haywire.
Any thoughts? If there was simply a way to determine the function's domain that would be helpful too. Thanks!
Oh, and also, the domain of the function is constricted. I don't know how to calculate the domain of the function either.
I've tried using a gradient search method already. I find the gradient at a point by calculating two points and doing it the old-fashioned way. Then I subtract the gradient from its point to get a new point. I keep searching this way until the gradient is relatively small (.000001). However, the code breaks a good deal because when the gradient is extremely large, this sends the calculated, new point out of the function's domain, and everything goes haywire.
Any thoughts? If there was simply a way to determine the function's domain that would be helpful too. Thanks!