![[2602.16316] A Graph Meta-Network for Learning on Kolmogorov-Arnold Networks](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.16316] A Graph Meta-Network for Learning on Kolmogorov-Arnold Networks
Summary
This paper introduces WS-KAN, a novel weight-space architecture for Kolmogorov-Arnold Networks (KANs), demonstrating its superior performance over traditional methods in various tasks.
Why It Matters
The development of WS-KAN addresses a significant gap in the design of architectures for KANs, which have unique permutation symmetries. This advancement can enhance machine learning applications, making it crucial for researchers and practitioners in the field.
Key Takeaways
- WS-KAN is the first architecture specifically designed for Kolmogorov-Arnold Networks.
- It leverages the permutation symmetries shared with MLPs to improve learning efficiency.
- Empirical evaluations show WS-KAN outperforms structure-agnostic baselines across diverse tasks.
- A comprehensive zoo of trained KANs serves as benchmarks for assessing WS-KAN's performance.
- The code for WS-KAN is publicly available, promoting further research and development.
Computer Science > Machine Learning arXiv:2602.16316 (cs) [Submitted on 18 Feb 2026] Title:A Graph Meta-Network for Learning on Kolmogorov-Arnold Networks Authors:Guy Bar-Shalom, Ami Tavory, Itay Evron, Maya Bechler-Speicher, Ido Guy, Haggai Maron View a PDF of the paper titled A Graph Meta-Network for Learning on Kolmogorov-Arnold Networks, by Guy Bar-Shalom and 5 other authors View PDF HTML (experimental) Abstract:Weight-space models learn directly from the parameters of neural networks, enabling tasks such as predicting their accuracy on new datasets. Naive methods -- like applying MLPs to flattened parameters -- perform poorly, making the design of better weight-space architectures a central challenge. While prior work leveraged permutation symmetries in standard networks to guide such designs, no analogous analysis or tailored architecture yet exists for Kolmogorov-Arnold Networks (KANs). In this work, we show that KANs share the same permutation symmetries as MLPs, and propose the KAN-graph, a graph representation of their computation. Building on this, we develop WS-KAN, the first weight-space architecture that learns on KANs, which naturally accounts for their symmetry. We analyze WS-KAN's expressive power, showing it can replicate an input KAN's forward pass - a standard approach for assessing expressiveness in weight-space architectures. We construct a comprehensive ``zoo'' of trained KANs spanning diverse tasks, which we use as benchmarks to empirically evaluate...