DREAM: Dual-Standard Semantic Homogeneity with Dynamic Optimization for Graph Learning with Label Noise

Graph neural networks (GNNs) have been widely used in various graph machine learning scenarios. Existing literature primarily assumes well-annotated training graphs, while the reliability of labels is not guaranteed in real-world scenarios. Recently, efforts have been made to address the problem of graph learning with label noise. However, existing methods often (i) struggle to distinguish between reliable and unreliable nodes, and (ii) overlook the relational information embedded in the graph topology. To tackle this problem, this paper proposes a novel method, Dual-Standard Semantic Homogeneity with Dynamic Optimization (DREAM), for reliable, relation-informed optimization on graphs with label noise. Specifically, we design a relation-informed dynamic optimization framework that iteratively reevaluates the reliability of each labeled node in the graph during the optimization process according to the relation of the target node and other nodes. To measure this relation comprehensively, we propose a dual-standard selection strategy that selects a set of anchor nodes based on both node proximity and graph topology. Subsequently, we compute the semantic homogeneity between the target node and the anchor nodes, which serves as guidance for optimization. We also provide a rigorous theoretical analysis to justify the design of DREAM. Extensive experiments are performed on six graph datasets across various domains under three types of graph label noise against competing baselines, and the results demonstrate the effectiveness of the proposed DREAM.

Machine learning models designed for networked data—such as social networks or biological structures—usually rely on perfectly accurate labels to learn effectively. However, in the real world, data is often mislabeled or "noisy." When faced with this noise, existing models struggle to figure out which data to trust, often ignoring the valuable connections between the data points themselves. To solve this, we introduce DREAM, a new method that helps AI learn reliably even from data with errors. Instead of blindly trusting the labels provided, our approach continuously double-checks the trustworthiness of each data point by looking at its surrounding neighborhood. We do this by strategically selecting highly reliable "reference points" based on two criteria: how closely related the data points are, and how they are structurally connected in the network. By comparing a data point to these trusted references, our model can naturally identify and filter out bad labels. We mathematically prove why this works and show through extensive testing that our method significantly outperforms current solutions across various types of noisy real-world data.