Online Continual Graph Learning

Giovanni Donghi, Luca Pasa, Daniele Zambon, Cesare Alippi, Nicolò Navarin

Keywords: continual learning, online learning, graph neural network

Abstract: The aim of Continual Learning (CL) is to learn new tasks incrementally while avoiding catastrophic forgetting. Online Continual Learning (OCL) specifically focuses on learning efficiently from a continuous stream of data with shifting distribution. While recent studies explore Continual Learning on graphs exploiting Graph Neural Networks (GNNs), only few of them focus on a streaming setting. Many real-world graphs evolve over time and timely (online) predictions could be required. However, current approaches are not well aligned with the standard OCL literature, partly due to the lack of a clear definition of online continual learning on graphs. In this work, we propose a general formulation for online continual learning on graphs, emphasizing the efficiency of batch processing while accounting for graph topology, providing a grounded setting to analyze different methods. We present a set of benchmark datasets for online continual graph learning, together with the results of several methods in CL literature, adapted to our setting. Additionally, we address the challenge of GNN memory usage, as considering multiple hops of neighborhood aggregation can require access to the entire growing graph, resulting in prohibitive costs for the setting. We thus propose solutions to maintain bounded complexity for efficient online learning.

Primary Area: transfer learning, meta learning, and lifelong learning

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.

Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Submission Number: 11334