search

Dlux804 / McQuade-Chem-ML

Development of easy to use and reproducible ML scripts for chemistry.
5 stars 1 forks source link

Sklearn classifiers failing with hyper-parameter tuning #84

Open dickeygh opened 4 years ago

dickeygh commented 4 years ago

Describe the bug SVM Classifier, linearSVC, and ada classifier raise errors when being ran with hyper-parameter tuning.

To Reproduce Run these models with tune=True in main.py.

Screenshots

SVM fail on tuning:

SVC fail on tuning part 1 SVC fail on tuning part 2

linearSVC fail on tuning: LinearSVC fail on tuning LinearSVC fail on tuning part 2

ada fail on tuning:

ada fail on tuning part 1 ada fail on tuning part 2

qle2 commented 4 years ago

@dickeygh For AdaboostClassifier, I think the error is caused by the hyperparameters. For regression, the base_estimator for AdaboostRegressor is DecisionTreeRegressor. So for classification, you should make another grid that uses DecisionTreeClassifier instead. I'm not sure about SVM though. Maybe you should look more into that on sklearn to see if you can find anything

Dlux804 commented 3 years ago

@dickeygh has this been fixed? If it has, please close.