decision_function instead of predict_proba #157

Activity

1. e-rich commented on Apr 28, 2022

   

   e-rich

   on Apr 28, 2022

   Last edited by e-rich

   More actions

   if you just want to use the closest-to-hyperplane selection with a linear SVM, you could also write yourself a query strategy like:

    def closest_to_hyperplane_sampling(linearSVC_learner: BaseLearner, X_pool):
       y = linearSVC_learner.estimator.decision_function(X_pool)
       w_norm = np.linalg.norm(linearSVC_learner.estimator.coef_)
       dist = y / w_norm
       query_idx = np.argmin(np.abs(dist))
       return query_idx, X_pool[query_idx]
   
2. lkurlandski commented on May 1, 2022

   

   lkurlandski

   on May 1, 2022

   Author

   More actions

   This is great, thank you for the tip. Did not realize how simple it would be.
3. lkurlandski commented on May 3, 2022

   

   lkurlandski

   on May 3, 2022

   Last edited by lkurlandski

   Author

   More actions

   I am curious why you normalize the confidence scores. I can't see how applying a linear scaling factor of (1 / w_norm) to the y vector would change the return value of the function. Is it simply to ensure all values are between 0 and 1?
4. e-rich commented on May 3, 2022

   

   e-rich

   on May 3, 2022

   More actions

   yes you are right. you do not need this step.
5. lkurlandski commented on May 3, 2022

   

   lkurlandski

   on May 3, 2022

   Author

   More actions

   Thought so. Cheers!