Classification from Positive, Unlabeled and Biased Negative Data
This work addresses a practical challenge in machine learning for scenarios where collecting fully representative negative data is difficult, offering an incremental improvement over existing PU learning methods.
The paper tackles the problem of binary classification when negative data are diverse and only a biased subset is available, by incorporating biased negative data into positive-unlabeled learning. The proposed method, based on empirical risk minimization and example weighting, shows effectiveness in both PUbN and ordinary PU learning scenarios on benchmark datasets.
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.