[SciPy-Dev] optimize - add algorithm for global optimization: GenSA

[Charles R Harris]

On Wed, Nov 18, 2015 at 11:23 PM, Sturla Molden <sturla.molden at gmail.com>
wrote:

> On 15/11/15 04:51, Charles R Harris wrote:
>
> The most reliable, if slow, method was a GA.
>>
>
> GAs are great, but it hard to wrap them in a general routine we could put
> in SciPy. The parameters have to be codable in such a way that we can
> meaningfully simulate recombinations and mutations.
>
> Tabu search is also a method that can compete with SA, but does not have
> the same tendency to go astray.
>

Might be interesting to discretize some of the test functions by rounding
either the function or its arguments. I expect step size would be critical
for algorithms heavily dependent on gradients.

Chuck
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/scipy-dev/attachments/20151119/7bd6deaf/attachment.html>