python - Early stopping in lgbm or xgb if score doesn't improve "enough" -
is there way, in xgboost or lgbm in python perform stopping if score doesn't improve more threshold?
i know early_stopping_rounds = n makes algorithm stop if score hasn't improved on evaluation set during n consecutive iterations, if score doesn't improve more fixed threshold.
i have learning task, in score on cv improves 0.65 0.25 in first 50 rounds (it's binary logloss) , continues on 1000 rounds achieve score of 0.241. rather stop when score doesn't improve more 0.001 in 10 rounds. there way that?
Comments
Post a Comment