Currently, HM does not start BO if all DoE samples are feasible in a constrained search space optimization problem. This is due to our classification model implementation for the feasibility constraint model. However, there is no reason not to start BO in that case, we can just not train the feasbility model until we hit an unfeasible point. We should change that.
Currently, HM does not start BO if all DoE samples are feasible in a constrained search space optimization problem. This is due to our classification model implementation for the feasibility constraint model. However, there is no reason not to start BO in that case, we can just not train the feasbility model until we hit an unfeasible point. We should change that.