Kolmogorov-Arnold Networks (KANs) for Time Series Analysis

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Vaca-Rubio, Cristian J., Blanco, Luis, Pereira, Roberto, Caus, Màrius
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915575052107776
author Vaca-Rubio, Cristian J.
Blanco, Luis
Pereira, Roberto
Caus, Màrius
author_facet Vaca-Rubio, Cristian J.
Blanco, Luis
Pereira, Roberto
Caus, Màrius
contents This paper introduces a novel application of Kolmogorov-Arnold Networks (KANs) to time series forecasting, leveraging their adaptive activation functions for enhanced predictive modeling. Inspired by the Kolmogorov-Arnold representation theorem, KANs replace traditional linear weights with spline-parametrized univariate functions, allowing them to learn activation patterns dynamically. We demonstrate that KANs outperforms conventional Multi-Layer Perceptrons (MLPs) in a real-world satellite traffic forecasting task, providing more accurate results with considerably fewer number of learnable parameters. We also provide an ablation study of KAN-specific parameters impact on performance. The proposed approach opens new avenues for adaptive forecasting models, emphasizing the potential of KANs as a powerful tool in predictive analytics.
format Preprint
id arxiv_https___arxiv_org_abs_2405_08790
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Kolmogorov-Arnold Networks (KANs) for Time Series Analysis
Vaca-Rubio, Cristian J.
Blanco, Luis
Pereira, Roberto
Caus, Màrius
Signal Processing
Artificial Intelligence
Machine Learning
This paper introduces a novel application of Kolmogorov-Arnold Networks (KANs) to time series forecasting, leveraging their adaptive activation functions for enhanced predictive modeling. Inspired by the Kolmogorov-Arnold representation theorem, KANs replace traditional linear weights with spline-parametrized univariate functions, allowing them to learn activation patterns dynamically. We demonstrate that KANs outperforms conventional Multi-Layer Perceptrons (MLPs) in a real-world satellite traffic forecasting task, providing more accurate results with considerably fewer number of learnable parameters. We also provide an ablation study of KAN-specific parameters impact on performance. The proposed approach opens new avenues for adaptive forecasting models, emphasizing the potential of KANs as a powerful tool in predictive analytics.
title Kolmogorov-Arnold Networks (KANs) for Time Series Analysis
topic Signal Processing
Artificial Intelligence
Machine Learning
url https://arxiv.org/abs/2405.08790