levelfour / SU_Classification

Implementation for "Classification from Pairwise Similarity and Unlabeled Data"

Classification from Pairwise Similarity and Unlabeled Data

This repository provides an official implementation of SU classification, which is a weakly-supervised classification problem only from pairwise similarity pairs (two data points belong to the same class) and unlabeled data points.

Dependencies

cvxopt==1.1.9
numpy==1.13.3
sklearn==0.18.1

Run

python su_learning.py --loss squared --ns 200 --nu 200 --prior 0.7

Notes

mpe.py is a (slightly-modified) implementation of mixture proportion estimation. We used the author's implementation available here.

References

• Bao, H., Niu, G., & Sugiyama, M. Classification from Pairwise Similarity and Unlabeled Data. In Proceedings of International Conference on Machine Learning (ICML), 2018. [arxiv]
• Ramaswamy, H. G., Scott, C., & Tewari, A. Mixture proportion estimation via kernel embedding of distributions. In Proceedings of International Conference on Machine Learning (ICML), pp. 2052–2060, 2016.