Deep Insights into Noisy Pseudo Labeling on Graph Data

TL;DR

This paper provides a theoretical analysis of pseudo labeling in graph learning, revealing how errors propagate and affect convergence, and proposes a cautious pseudo labeling method that enhances performance on graph tasks.

Contribution

It offers the first theoretical error analysis of pseudo labeling on graphs and introduces a confidence-based cautious pseudo labeling strategy.

Findings

The error in pseudo labeling is bounded by confidence and multi-view consistency.

The proposed cautious pseudo labeling improves performance on link prediction and node classification.

The strategy outperforms existing pseudo labeling methods in experiments.

Abstract

Pseudo labeling (PL) is a wide-applied strategy to enlarge the labeled dataset by self-annotating the potential samples during the training process. Several works have shown that it can improve the graph learning model performance in general. However, we notice that the incorrect labels can be fatal to the graph training process. Inappropriate PL may result in the performance degrading, especially on graph data where the noise can propagate. Surprisingly, the corresponding error is seldom theoretically analyzed in the literature. In this paper, we aim to give deep insights of PL on graph learning models. We first present the error analysis of PL strategy by showing that the error is bounded by the confidence of PL threshold and consistency of multi-view prediction. Then, we theoretically illustrate the effect of PL on convergence property. Based on the analysis, we propose a cautious pseudo labeling methodology in which we pseudo label the samples with highest confidence and multi-view consistency. Finally, extensive experiments demonstrate that the proposed strategy improves graph learning process and outperforms other PL strategies on link prediction and node classification tasks.