Time series forecasting with Hahn Kolmogorov-Arnold networks

TL;DR

The paper introduces HaKAN, a lightweight, interpretable model based on Kolmogorov-Arnold Networks with Hahn polynomial activations, achieving superior long-term multivariate time series forecasting performance.

Contribution

It presents HaKAN, a novel model combining KANs and Hahn polynomials, offering an efficient, interpretable alternative to Transformers and MLPs for time series forecasting.

Findings

Outperforms state-of-the-art forecasting models on multiple benchmarks.

Effective capture of global and local temporal patterns.

Ablation studies confirm the importance of core components.

Abstract

Recent Transformer- and MLP-based models have demonstrated strong performance in long-term time series forecasting, yet Transformers remain limited by their quadratic complexity and permutation-equivariant attention, while MLPs exhibit spectral bias. We propose HaKAN, a versatile model based on Kolmogorov-Arnold Networks (KANs), leveraging Hahn polynomial-based learnable activation functions and providing a lightweight and interpretable alternative for multivariate time series forecasting. Our model integrates channel independence, patching, a stack of Hahn-KAN blocks with residual connections, and a bottleneck structure comprised of two fully connected layers. The Hahn-KAN block consists of inter- and intra-patch KAN layers to effectively capture both global and local temporal patterns. Extensive experiments on various forecasting benchmarks demonstrate that our model consistently outperforms recent state-of-the-art methods, with ablation studies validating the effectiveness of its core components.