I've been musing on the implementation of non-linear constraints for `scipy.optimize.differential_evolution`. There are various ways of doing this (search for "nonlinear constraint function differential evolution" on Google). There is a version by Lampinen (2002) that seems fairly straightforward. It is described at: https://pdfs.semanticscholar.org/088e/60df2694230e8e9e841e19ec218cddba54fe.pdf

The paper has been cited 263 times, which indicates its usefulness. These kinds of constraints have been implemented in the R version (as indicated by the documentation).

Does anyone else have any thoughts as to whether they'd find such an approach useful in scipy?

Andrew.

--