Hi @keurfonluu, thanks for this great library!
If I look at the code, I understand that the bounds parameter is only used to initialize the population of parameters vectors.
Would it be possible to actually enforce the bounds during optimization to limit the search space? I know I can artificially enforce it in the cost function, but it means that the cost function does not really corresponds to the parameters values sent by the optimizer.
It is also doable thanks to a variable change but it also means that lower and upper bounds are not reachable (asymptotic behaviour at the bounds)
Do you have an opinion about that ?
If you think this could be interesting, I guess a simple Boolean such as is_enforce_bounds and a bit of clipping in the code would be enough.
Best
Antoine
Hi @keurfonluu, thanks for this great library!
If I look at the code, I understand that the
boundsparameter is only used to initialize the population of parameters vectors.Would it be possible to actually enforce the bounds during optimization to limit the search space? I know I can artificially enforce it in the cost function, but it means that the cost function does not really corresponds to the parameters values sent by the optimizer.
It is also doable thanks to a variable change but it also means that lower and upper bounds are not reachable (asymptotic behaviour at the bounds)
Do you have an opinion about that ?
If you think this could be interesting, I guess a simple Boolean such as
is_enforce_boundsand a bit of clipping in the code would be enough.Best
Antoine