Kolmogorov-Arnold Networks (KANs) for Time Series Analysis

TL;DR

This paper presents Kolmogorov-Arnold Networks (KANs), a novel neural network architecture for time series forecasting that uses adaptive activation functions inspired by the Kolmogorov-Arnold theorem, outperforming traditional models.

Contribution

Introduces KANs with spline-parametrized activation functions for improved time series prediction, demonstrating superior performance over MLPs with fewer parameters.

Findings

KANs outperform MLPs in satellite traffic forecasting

KANs require fewer parameters for comparable accuracy

Ablation study highlights importance of KAN-specific parameters

Abstract

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.