Some Sklearn scoring functions return negative error scores (e.g. mean absolute error) to stick to the "bigger is better" paradigm. However, with this error score the following code block never is evaluated True
if current_best_score_ > self.best_score_: self.best_score_ = current_best_score_ self.best_params_ = current_best_params_
and the best_score, and more importantly the best_params stay -1 and None respectively.
Setting self.bestscore with -inf or with a much smaller value would sort of solve the issue. Setting best_score with current_best_score in the first generation would be a better solution imo
rsteca
commented
8 years ago
Some Sklearn scoring functions return negative error scores (e.g. mean absolute error) to stick to the "bigger is better" paradigm. However, with this error score the following code block never is evaluated True
if current_best_score_ > self.best_score_: self.best_score_ = current_best_score_ self.best_params_ = current_best_params_and the best_score, and more importantly the best_params stay -1 and None respectively.
Setting self.bestscore with -inf or with a much smaller value would sort of solve the issue. Setting best_score with current_best_score in the first generation would be a better solution imo