🤖 AI Summary
This paper identifies “excessive aggregation” as the root cause of oversmoothing in Graph Convolutional Networks (GCNs): higher-order neighborhood messages must propagate layer-by-layer through lower-order neighbors, leading to redundant and repeated information aggregation. To address this, we propose a redundancy-aware neighbor aggregation paradigm that abandons conventional deep stacking architectures. Specifically, we design a single-layer aggregator grounded in adjacency matrix power analysis and neighbor contribution modeling. Our approach eliminates the need for residual connections or stochastic edge dropout. Empirically, it significantly alleviates oversmoothing, improves performance of deep GCNs across multiple benchmark datasets, accelerates convergence, enhances generalization, and reduces model parameters by approximately 40%. The method is both structurally simple and highly scalable.
📝 Abstract
In recent years, Graph Convolutional Networks (GCNs) have gained popularity for their exceptional ability to process graph-structured data. Existing GCN-based approaches typically employ a shallow model architecture due to the over-smoothing phenomenon. Current approaches to mitigating over-smoothing primarily involve adding supplementary components to GCN architectures, such as residual connections and random edge-dropping strategies. However, these improvements toward deep GCNs have achieved only limited success. In this work, we analyze the intrinsic message passing mechanism of GCNs and identify a critical issue: messages originating from high-order neighbors must traverse through low-order neighbors to reach the target node. This repeated reliance on low-order neighbors leads to redundant information aggregation, a phenomenon we term over-aggregation. Our analysis demonstrates that over-aggregation not only introduces significant redundancy but also serves as the fundamental cause of over-smoothing in GCNs.