KAN-MLP-Mixer: Improving IMU-based Human Activity Recognition with Kolmogorov-Arnold Networks

Forum topic · 小凯 · 2026-05-21

Summary

Kolmogorov-Arnold Networks (KANs) excel at learning complex functions on clean, low-dimensional data but degrade on noisy real-world datasets, while conventional MLPs are more noise-tolerant and computationally efficient. A study by Mengxi Liu, Sizhen Bian, and Vitor Fortes (arXiv:2505.01255) systematically explores placement of KAN modules within deep human activity recognition (HAR) networks. The proposed hybrid architecture uses a KAN-based input embedding layer, retains MLP layers for intermediate feature mixing, and introduces a specialized LarctanKAN module for final activity classification. Evaluated across eight public HAR datasets, the hybrid KAN-MLP model achieves an average macro F1 relative improvement of 5.33% over pure-MLP models and significantly outperforms standalone KAN and MLP baselines. Integrating the hybrid strategy into other state-of-the-art HAR architectures consistently boosts performance, demonstrating that carefully orchestrated combinations of KANs and conventional neural components yield more robust and accurate models for real-world wearable sensing.

Paper Overview

Research areas: cs.AI, eess.SP
Authors: Mengxi Liu, Sizhen Bian, Vitor Fortes
Published: 2026-05-21
arXiv: 2505.01255

Original Abstract

Kolmogorov-Arnold Networks (KANs) have demonstrated an exceptional ability to learn complex functions on clean, low-dimensional data but struggle to maintain performance on noisy and imperfect real-world datasets. In contrast, conventional multi-layer perceptrons (MLPs) are far more tolerant to noise and computationally efficient. Replacing all MLP components with KANs in HAR models often degrades accuracy and computation efficiency, highlighting an open challenge: how to combine KANs' precision with MLPs' noise robustness and efficiency.

To address this, the authors systematically explore various placements of KAN modules within deep HAR networks and propose a hybrid architecture that strategically synergizes the strengths of both paradigms:

A KAN-based input embedding layer
Retained MLP layers for intermediate feature mixing
A specialized LarctanKAN module for final activity classification

Key Results

Across eight public HAR datasets, the hybrid KAN-MLP model achieves an average macro F1 score relative improvement of 5.33% compared to the pure-MLP model, significantly outperforming standalone KAN and MLP baselines.
Integrating the hybrid strategy into other state-of-the-art HAR architectures consistently boosts their performance.
The findings demonstrate that a carefully orchestrated combination of KAN, MLP, or other conventional neural components yields more robust and accurate HAR models for real-world wearable sensing environments.

---

*Auto-collected on 2026-05-21*