scipy - Python sparse matrix & classifiers -
working naturally sparse data love use scipy sparse type (csr_matrix instance). confused when @ various classifiers on whether support type or not !
- logistic regression, naive bayes seem working
- as regards svm classifiers (sklearn.svm.svc instance) understand can predict on same type has first fitted (so if fit on sparse matrix, can use .predict() , .score() sparse data), correct ?
a second question pros of using sparse type (appart memory storage) : improving speed of algorithm ? svm.svc doesn't so, unable tell sure, know more on subject ?
i have add planning on using cross_val_score , greadsearchcv tune , cross-validate on sparse data, , love be sure functions can support efficiently sparse type.
thank !
Comments
Post a Comment