Classification from Pairwise Similarity and Unlabeled Data
This addresses the bottleneck of high labeling costs or privacy concerns in supervised learning, though it is an incremental improvement over existing weakly-supervised methods.
The paper tackles the problem of reducing the need for labeled data in supervised learning by proposing a weakly-supervised setting called SU classification, which uses only similar data pairs and unlabeled data, and shows that this method achieves optimal parametric convergence rates in experiments.
Supervised learning needs a huge amount of labeled data, which can be a big bottleneck under the situation where there is a privacy concern or labeling cost is high. To overcome this problem, we propose a new weakly-supervised learning setting where only similar (S) data pairs (two examples belong to the same class) and unlabeled (U) data points are needed instead of fully labeled data, which is called SU classification. We show that an unbiased estimator of the classification risk can be obtained only from SU data, and the estimation error of its empirical risk minimizer achieves the optimal parametric convergence rate. Finally, we demonstrate the effectiveness of the proposed method through experiments.