SigKAN: Signature-Weighted Kolmogorov-Arnold Networks for Time Series

Hugo Inzirillo, Remi Genet
6/25/2024
cs.LG

No official code found. Best implementation: remigenet/SigKAN (49 stars, last commit 2024-11-24).

Abstract

We propose a novel approach that enhances multivariate function approximation using learnable path signatures and Kolmogorov-Arnold networks (KANs). We enhance the learning capabilities of these networks by weighting the values obtained by KANs using learnable path signatures, which capture important geometric features of paths. This combination allows for a more comprehensive and flexible representation of sequential and temporal data. We demonstrate through studies that our SigKANs with learnable path signatures perform better than conventional methods across a range of function approximation challenges. By leveraging path signatures in neural networks, this method offers intriguing opportunities to enhance performance in time series analysis and time series forecasting, among other fields.

View on arXivView PDF

Code Implementations(1)

remigenet/SigKANCommunity100%

SigKAN: Signature-Weighted Kolmogorov-Arnold Networks for Time Series

4913PythonJun 15, 20241 years agoNOASSERTION
kolmogorov-arnold-networksrough-pathssignaturetemporaltensorflow2+1 more

Similar Papers

Kolmogorov-Arnold Networks (KANs) for Time Series Analysis

87% match2024

Kolmogorov-Arnold Networks for Time Series: Bridging Predictive Power and Interpretability

87% match2024

A Survey on Kolmogorov-Arnold Network

86% match2024

Kolmogorov-Arnold Networks (KAN) for Time Series Classification and Robust Analysis

85% match2024

KKANs: Kurkova-Kolmogorov-Arnold Networks and Their Learning Dynamics

85% match2024

TSKANMixer: Kolmogorov-Arnold Networks with MLP-Mixer Model for Time Series Forecasting

84% match2025

Cite this paper

@article{inzirillo2024sigkan,
  title  = {SigKAN: Signature-Weighted Kolmogorov-Arnold Networks for Time Series},
  author = {Hugo Inzirillo and Remi Genet},
  year   = {2024},
  eprint = {2406.17890},
  archivePrefix = {arXiv},
  url    = {http://arxiv.org/abs/2406.17890v2}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2406.17890.svg)](https://codeofpaper.com/paper/2406.17890)

Discussion