Weak Supervision for Real World Graphs

TL;DR

This paper introduces WSNET, a weakly supervised graph contrastive learning framework that effectively utilizes noisy and indirect signals for node classification in real-world graphs with limited labels.

Contribution

The paper presents WSNET, a novel contrastive learning method that leverages weak signals and noisy supervision sources for improved node classification in challenging graph settings.

Findings

WSNET outperforms existing methods by up to 15% in F1 score.

Contrastive learning effectively utilizes weak and noisy supervision.

WSNET demonstrates robustness across multiple real-world datasets.

Abstract

Node classification in real world graphs often suffers from label scarcity and noise, especially in high stakes domains like human trafficking detection and misinformation monitoring. While direct supervision is limited, such graphs frequently contain weak signals, noisy or indirect cues, that can still inform learning. We propose WSNET, a novel weakly supervised graph contrastive learning framework that leverages these weak signals to guide robust representation learning. WSNET integrates graph structure, node features, and multiple noisy supervision sources through a contrastive objective tailored for weakly labeled data. Across three real world datasets and synthetic benchmarks with controlled noise, WSNET consistently outperforms state of the art contrastive and noisy label learning methods by up to 15% in F1 score. Our results highlight the effectiveness of contrastive learning under weak supervision and the promise of exploiting imperfect labels in graph based settings.