![[2602.15711] Random Wavelet Features for Graph Kernel Machines](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.15711] Random Wavelet Features for Graph Kernel Machines
Summary
This paper introduces randomized spectral node embeddings for graph kernel machines, enhancing node similarity estimation while improving computational efficiency for large networks.
Why It Matters
As graph-based data becomes increasingly prevalent in machine learning, developing efficient methods for node representation is crucial. This research offers a scalable approach to graph kernel approximation, which can significantly impact various applications like social network analysis and bioinformatics.
Key Takeaways
- Randomized spectral node embeddings provide an efficient way to approximate graph kernels.
- The proposed method achieves better accuracy than existing techniques, especially for spectrally localized kernels.
- This approach addresses scalability issues in graph representation learning.
Computer Science > Machine Learning arXiv:2602.15711 (cs) [Submitted on 17 Feb 2026] Title:Random Wavelet Features for Graph Kernel Machines Authors:Valentin de Bassompierre, Jean-Charles Delvenne, Laurent Jacques View a PDF of the paper titled Random Wavelet Features for Graph Kernel Machines, by Valentin de Bassompierre and 1 other authors View PDF HTML (experimental) Abstract:Node embeddings map graph vertices into low-dimensional Euclidean spaces while preserving structural information. They are central to tasks such as node classification, link prediction, and signal reconstruction. A key goal is to design node embeddings whose dot products capture meaningful notions of node similarity induced by the graph. Graph kernels offer a principled way to define such similarities, but their direct computation is often prohibitive for large networks. Inspired by random feature methods for kernel approximation in Euclidean spaces, we introduce randomized spectral node embeddings whose dot products estimate a low-rank approximation of any specific graph kernel. We provide theoretical and empirical results showing that our embeddings achieve more accurate kernel approximations than existing methods, particularly for spectrally localized kernels. These results demonstrate the effectiveness of randomized spectral constructions for scalable and principled graph representation learning. Comments: Subjects: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Signal Processing (e...
