Open RockyGHub opened 9 months ago
Hi @RockyGHub,
I will have a look at this.
Just one comment: for historical data which has acutally be done (experiments not candidates), you should use validate_experiments
, as it is not enforced that the historical data fullfills the constraints which means that you can have training data for the surrogates that does not fulfill the constraints.
Best,
Johannes
I'd have a wish to get an additional information which constrain stays in conflict to historical data.
The function domain.validate_candidates gives nice outputs if inputs are outside of the defined range e.g.
B10
are smaller than upper bound0.8
In case of conflict with constrains were is no detailed information which constrain is violated.
Now, I have to switch on and off all constrains to understand where the problem is. The desired error message would be 'Constrain 1 : ..... is violated '
TnX!