Unified risk analysis for weakly supervised learning

CK Chiang, M Sugiyama - arxiv preprint arxiv:2309.08216, 2023 - arxiv.org
Among the flourishing research of weakly supervised learning (WSL), we recognize the lack
of a unified interpretation of the mechanism behind the weakly supervised scenarios, let …

Learning with proper partial labels

Z Wu, J Lv, M Sugiyama - Neural Computation, 2023 - direct.mit.edu
Partial-label learning is a kind of weakly supervised learning with inexact labels, where for
each training example, we are given a set of candidate labels instead of only one true label …

Pairwise Confidence Difference on Unlabeled Data is Sufficient for Binary Classification

W Wang, L Feng, G Niu, ML Zhang, M Sugiyama - openreview.net
Learning with confidence labels is an emerging weakly supervised learning paradigm,
where training data are equipped with confidence labels instead of exact labels. Positive …

[PDF][PDF] Binary Classification from Uncertainty and Triplet Comparison

Z Cui - 2021 - repository.dl.itc.u-tokyo.ac.jp
Machine learning has achieved tremendous development and brought innovation to many
aspects of the society in the recent decade. Binary classification, one of the core tasks of …