Classification from Pairwise Similarity and Unlabeled Data

Han Bao, Gang Niu, Masashi Sugiyama
2/12/2018
cs.LG

Official code: levelfour/SU_Classification (27 stars, MIT, last commit 2020-05-12).

Abstract

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.

View on arXivView PDF

Code Implementations(1)

levelfour/SU_ClassificationOfficial100%
MIT

Implementation for "Classification from Pairwise Similarity and Unlabeled Data"

277PythonJun 3, 20186 years agoMIT

Similar Papers

Learning from Uncertain Similarity and Unlabeled Data

86% match1 repo2025

Multi-class Classification from Multiple Unlabeled Datasets with Partial Risk Regularization

84% match2022

MixMatch: A Holistic Approach to Semi-Supervised Learning

84% match1 repo1.1k2019

Learning from Reduced Labels for Long-Tailed Data

83% match1 repo2024

FlatMatch: Bridging Labeled Data and Unlabeled Data with Cross-Sharpness for Semi-Supervised Learning

83% match1 repo82023

Improving 3D Semi-supervised Learning by Effectively Utilizing All Unlabelled Data

83% match1 repo52024

Cite this paper

@article{bao2018classification,
  title  = {Classification from Pairwise Similarity and Unlabeled Data},
  author = {Han Bao and Gang Niu and Masashi Sugiyama},
  year   = {2018},
  eprint = {1802.04381},
  archivePrefix = {arXiv},
  url    = {https://arxiv.org/abs/1802.04381}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/1802.04381.svg)](https://codeofpaper.com/paper/1802.04381)

Discussion