enh: aggressive DE strategies

[pavelponomarev]

Hi,
Is it possible to add some aggressive strategies to DE like "current-to-best/1/bin"?
http://www.dii.unipd.it/~alotto/didattica/corsi/Elettrotecnica%20computazionale/DE.pdf
BR, Pavel

[dlax]

Sure, patch welcome ;)

[pavelponomarev]

After more thorough code examination, and after discussions in #4864 , it seems that implemented DE algorithm in scipy is already aggressive -- it can mix current generation with the best members from itself, so there is no separation between generations. So, this information about aggressiveness of the mutation strategy must be added to the documentation of the module.
So, actually I want to do exactly the opposite right now) - implement possibility for non aggressive mutation strategy, which is more suitable for parallelization. The works on that are under discussion in #4864.

[argriffing]

After more thorough code examination, and after discussions in #4864 , it seems that implemented DE algorithm in scipy is already aggressive

I'll close this issue then.

[argriffing]

information about aggressiveness of the mutation strategy must be added to the documentation of the module.

@pavelponomarev do you want to make a PR to add this clarification? What about just mentioning in the docstring that the default strategy best1bin is relatively aggressive and uses overlapping generations (if this is true)?

[pavelponomarev]

@argriffing let's wait results of the discussion in #4864