search

xopt-org / Xopt

Flexible high-level optimization in Python
Apache License 2.0
71 stars 24 forks source link

TuRBO refuses to work if no evaluated points are feasible #217

Closed wenatuhs closed 7 months ago

wenatuhs commented 7 months ago

Encountered error from TuRBO w/ constrained optimization – need at least 1 valid point in the training dataset.

When the constraints are conservative, it's quite possible that the initial current and random points are not feasible solutions, in this case, TuRBO refuses to sample more points to break the tie.

roussel-ryan commented 7 months ago

this is correct behavior, we could re-open this if a strong case comes up down the road