I am trying to do a grid searching using the methodology that mentioned in this post. However, I found that
XGBClassifier().fit() is using much more memory than
xgboost.train. Does anyone know why? Is this related to sparse matrix?
XGBRegressor vs. xgboost.train huge speed difference? - see this answer, likely the defaults (num_boost_round=10 vs. n_estimators=100) are the cause of this.