Support for constrained optimization

[miha-skalic]

Hello there,

first of all, I would like to thank you for the package.

Secondly, I have a request. Many real-world optimization problems of interest have constraints which are unknown a priori. It would be very cool if this package could implement constrained Bayesian optimization as reported in this paper: https://arxiv.org/abs/1403.5607

Essentially, one would define probabilistic constraints of the sort:

And place a GP surrogate on each constraint. These are updated in each evaluation of the target function as well.

Best, Miha

[josejimenezluna]

Hi,

This functionality was also suggested by https://www.reddit.com/r/MachineLearning/comments/6e6x8c/p_pygpgo_another_python_package_for_bayesian/di9ct52/

I will consider adding it in the near future.

[jacktang]

Hello @josejimenezluna , since the issue was marked as completed, can you provide one example to show how to add constraints?