Our new X account is live! Follow @wizwand_team for updates
WorkDL logo mark

Classification from Positive, Unlabeled and Biased Negative Data

About

In binary classification, there are situations where negative (N) data are too diverse to be fully labeled and we often resort to positive-unlabeled (PU) learning in these scenarios. However, collecting a non-representative N set that contains only a small portion of all possible N data can often be much easier in practice. This paper studies a novel classification framework which incorporates such biased N (bN) data in PU learning. We provide a method based on empirical risk minimization to address this PUbN classification problem. Our approach can be regarded as a novel example-weighting algorithm, with the weight of each example computed through a preliminary step that draws inspiration from PU learning. We also derive an estimation error bound for the proposed method. Experimental results demonstrate the effectiveness of our algorithm in not only PUbN learning scenarios but also ordinary PU learning scenarios on several benchmark datasets.

Yu-Guan Hsieh, Gang Niu, Masashi Sugiyama• 2018

Related benchmarks

TaskDatasetResultRank
Positive-Unlabeled ClassificationCIFAR-10 (test)
Accuracy89.83
19
Positive-Unlabeled LearningSVHN (test)
Accuracy0.8489
15
Positive-Unlabeled LearningSTL-10 (test)
Accuracy94.01
14
Positive-Unlabeled ClassificationAlzheimer dataset
F1 Score70.4
11
Positive-Unlabeled LearningADNI (test)
Accuracy0.7
6
Showing 5 of 5 rows

Other info

Follow for update