martinus / differential-evolution-rs Goto Github PK

You may want to consider Rust Random Choice for updating your samples. It can handle directly float values and it is simpler to use than the rand crate.

But the crate scores the costs diffenrently: The higher, the better. So you have to do this:

let fitness_high = MAX_VALUE - fitness_low;

Hello Martinus,

Great work on this crate. It is very fast... much faster than code I have written using 'Scipy minimize' in Python. I did notice though, it is rather unusable if we have constrained input parameters. It seems as though we pass in an initial min-max bound for each variable, but after that it is disregarded. Is there a way to keep these variables bounded? Or if not, can you point me to the section of code where this bound-breaking operation occurs? I would like to update it so I can use this great work you have done.

-Chase