DeGLIF for Label Noise Robust Node Classification using GNNs
This addresses the problem of noisy labels in graph datasets for researchers and practitioners using GNNs, offering a method that does not require noise model information, though it is incremental as it builds on recent advances in influence functions for GNNs.
The paper tackles label noise in graph data for node classification by proposing DeGLIF, a denoising technique that uses a small clean set and leave-one-out influence functions to robustly predict labels, achieving better accuracy than baseline algorithms in experiments.
Noisy labelled datasets are generally inexpensive compared to clean labelled datasets, and the same is true for graph data. In this paper, we propose a denoising technique DeGLIF: Denoising Graph Data using Leave-One-Out Influence Function. DeGLIF uses a small set of clean data and the leave-one-out influence function to make label noise robust node-level prediction on graph data. Leave-one-out influence function approximates the change in the model parameters if a training point is removed from the training dataset. Recent advances propose a way to calculate the leave-one-out influence function for Graph Neural Networks (GNNs). We extend that recent work to estimate the change in validation loss, if a training node is removed from the training dataset. We use this estimate and a new theoretically motivated relabelling function to denoise the training dataset. We propose two DeGLIF variants to identify noisy nodes. Both these variants do not require any information about the noise model or the noise level in the dataset; DeGLIF also does not estimate these quantities. For one of these variants, we prove that the noisy points detected can indeed increase risk. We carry out detailed computational experiments on different datasets to show the effectiveness of DeGLIF. It achieves better accuracy than other baseline algorithms