Remove hyperparameter optimization + useless optimizers
Closed this issue · 1 comments
ardunn commented
Doing hyperparameter optimization on RF and GP at every iteration seems to perform worse than random on average. There is not much need for hyperparameter optimization of the regressor if it is both slower and worse performance
Similarly, there are other regressors that simply don't work out of the box or with few data points (ie, MLPRegressor), which should be removed from the default options