![[2602.13634] Optimization-Free Graph Embedding via Distributional Kernel for Community Detection](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.13634] Optimization-Free Graph Embedding via Distributional Kernel for Community Detection
Summary
This article presents a novel method for graph embedding that addresses over-smoothing in Neighborhood Aggregation Strategy (NAS) methods, enhancing community detection without requiring optimization.
Why It Matters
The findings are significant as they tackle a common limitation in graph embedding techniques, which can hinder the performance of machine learning models in community detection tasks. By introducing a distribution-aware kernel, the research offers a new approach that maintains node distinguishability and improves the effectiveness of existing methods.
Key Takeaways
- Introduces a weighted distribution-aware kernel for graph embedding.
- Addresses the over-smoothing issue prevalent in NAS methods.
- No optimization is required, simplifying the embedding process.
- Demonstrates superior performance in community detection via spectral clustering.
- Incorporates critical node distribution characteristics often overlooked.
Computer Science > Machine Learning arXiv:2602.13634 (cs) [Submitted on 14 Feb 2026] Title:Optimization-Free Graph Embedding via Distributional Kernel for Community Detection Authors:Shuaibin Song, Kai Ming Ting, Kaifeng Zhang, Tianrun Liang View a PDF of the paper titled Optimization-Free Graph Embedding via Distributional Kernel for Community Detection, by Shuaibin Song and 3 other authors View PDF HTML (experimental) Abstract:Neighborhood Aggregation Strategy (NAS) is a widely used approach in graph embedding, underpinning both Graph Neural Networks (GNNs) and Weisfeiler-Lehman (WL) methods. However, NAS-based methods are identified to be prone to over-smoothing-the loss of node distinguishability with increased iterations-thereby limiting their effectiveness. This paper identifies two characteristics in a network, i.e., the distributions of nodes and node degrees that are critical for expressive representation but have been overlooked in existing methods. We show that these overlooked characteristics contribute significantly to over-smoothing of NAS-methods. To address this, we propose a novel weighted distribution-aware kernel that embeds nodes while taking their distributional characteristics into consideration. Our method has three distinguishing features: (1) it is the first method to explicitly incorporate both distributional characteristics; (2) it requires no optimization; and (3) it effectively mitigates the adverse effects of over-smoothing, allowing WL to pre...
