IJCNN 2025 Competition: Learning with Noisy Graph Labels

Farooq Ahmad Wani, Maria Sofia Bucarelli, Giulia Di Teodoro, Andrea Giuseppe Di Francesco

Abstract: Handling noisy data is a longstanding challenge in machine learning, and the complexity increases when working with graph structured data. In domains such as social networks, biological networks, and financial systems, noisy labels can significantly impact model performance, leading to unreliable predictions. Despite its importance, graph classification under label noise remains a relatively underexplored problem. In this competition, Learning with Noisy Graph Labels, we invite researchers and practitioners to develop noise robust approaches for graph classification. By fostering the development of robust models, it enhances the reliability of graph based predictions in real world scenarios where obtaining clean labels is difficult. The competition also promotes broader adoption of graph learning in industry by providing more resilient decision making tools for noisy labeled data. In addition, it contributes to benchmarking and standardization, offering a baseline data set and evaluation protocol to drive future research in noise resistant graph classification.

External IDs:dblp:conf/ijcnn/WaniBTF25