ADMP-GNN: Adaptive Depth Message Passing GNN

Yassine ABBAHADDOU, Fragkiskos D. Malliaros, Johannes F. Lutzeyer, Michalis Vazirgiannis

Keywords: Graph Neural Networks

TL;DR: we propose Adaptive Depth Message Passing GNN (ADMP-GNN), a novel framework that dynamically adjusts the number of message-passing layers for each node, leading to enhanced performance.

Abstract: Graph Neural Networks (GNNs) have proven to be highly effective in various graph representation learning tasks. A key characteristic is that GNNs apply a fixed number of message-passing steps to all nodes in the graph, regardless of the varying computational needs and characteristics of each node. Through empirical analysis of real-world data, we show that the optimal number of message-passing layers differs for nodes with different characteristics. This insight is further validated with experiments on synthetic datasets. To address this, we propose Adaptive Depth Message Passing GNN (ADMP-GNN), a novel framework that dynamically adjusts the number of message-passing layers for each node, leading to enhanced performance. This approach is applicable to any model that follows the message-passing scheme. We evaluate ADMP-GNN on the node classification task and observe performance improvements over a wide range of GNNs.

Primary Area: learning on graphs and other geometries & topologies

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.

Reciprocal Reviewing: I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6.

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: 10838