Masked Label Prediction: Unified Message Passing Model for Semi-Supervised Classification

TLDR

Graph neural networks and label propagation algorithms are powerful message‑passing methods for semi‑supervised classification, yet no effective approach exists to combine them. The authors introduce UniMP, a unified message‑passing model that integrates feature and label propagation during both training and inference. UniMP employs a Graph Transformer that takes feature and label embeddings as input and uses a masked‑label prediction strategy to prevent overfitting to self‑loop label information. Empirically, UniMP unifies the two propagation types and achieves new state‑of‑the‑art semi‑supervised classification performance on the Open Graph Benchmark.

Abstract

Graph neural network (GNN) and label propagation algorithm (LPA) are both message passing algorithms, which have achieved superior performance in semi-supervised classification. GNN performs feature propagation by a neural network to make predictions, while LPA uses label propagation across graph adjacency matrix to get results. However, there is still no effective way to directly combine these two kinds of algorithms. To address this issue, we propose a novel Unified Message Passaging Model (UniMP) that can incorporate feature and label propagation at both training and inference time. First, UniMP adopts a Graph Transformer network, taking feature embedding and label embedding as input information for propagation. Second, to train the network without overfitting in self-loop input label information, UniMP introduces a masked label prediction strategy, in which some percentage of input label information are masked at random, and then predicted. UniMP conceptually unifies feature propagation and label propagation and is empirically powerful. It obtains new state-of-the-art semi-supervised classification results in Open Graph Benchmark (OGB).