search

CMA-ES / libcmaes

libcmaes is a multithreaded C++11 library with Python bindings for high performance blackbox stochastic optimization using the CMA-ES algorithm for Covariance Matrix Adaptation Evolution Strategy
Other
321 stars 78 forks source link

Custom Step Size #232

Open godrays opened 2 years ago

godrays commented 2 years ago

Hello,

I'm just wondering if the custom step size is possible to use. I believe this would optimize the entire process.

Assume that some parameters can only be an integer value, although they are floating-point values in the process. For example, a x parameter is meaningful only when its values are 0,1,2,3,4,5,6 .... step size (1) OR another parameter's value can only be 0, 0.5, 1, 1.5, 2, ... etc.. step size (0.5).

Please excuse me if I miss something in the documentation.

nikohansen commented 2 years ago

To make this work is a little more complicated than you might think. This is a reference that tries to address this problem: https://hal.inria.fr/file/index/docid/629689/filename/RR-7751.pdf I think this has only been implemented in Octave/Matlab.

godrays commented 2 years ago

Nice! Any reason why you haven't implemented it in this library?