Label Propagation with Weak Supervision

TL;DR

This paper enhances label propagation algorithms by integrating prior probabilistic labels and multiple noisy sources, providing theoretical error bounds and demonstrating improved performance on weakly supervised classification benchmarks.

Contribution

It introduces a novel analysis of label propagation that leverages prior information and multiple noisy sources, advancing semi-supervised learning methods.

Findings

Improved classification accuracy on benchmark weakly supervised tasks.

Theoretical error bounds based on graph geometry and prior quality.

Effective incorporation of multiple noisy label sources.

Abstract

Semi-supervised learning and weakly supervised learning are important paradigms that aim to reduce the growing demand for labeled data in current machine learning applications. In this paper, we introduce a novel analysis of the classical label propagation algorithm (LPA) (Zhu & Ghahramani, 2002) that moreover takes advantage of useful prior information, specifically probabilistic hypothesized labels on the unlabeled data. We provide an error bound that exploits both the local geometric properties of the underlying graph and the quality of the prior information. We also propose a framework to incorporate multiple sources of noisy information. In particular, we consider the setting of weak supervision, where our sources of information are weak labelers. We demonstrate the ability of our approach on multiple benchmark weakly supervised classification tasks, showing improvements upon existing semi-supervised and weakly supervised methods.