Exploring Regularization Methods for Domain Generalization in Accelerometer-Based Human Activity Recognition.

Autor: Bento N; Associação Fraunhofer Portugal Research, Rua Alfredo Allen 455/461, 4200-135 Porto, Portugal., Rebelo J; Associação Fraunhofer Portugal Research, Rua Alfredo Allen 455/461, 4200-135 Porto, Portugal., Carreiro AV; Associação Fraunhofer Portugal Research, Rua Alfredo Allen 455/461, 4200-135 Porto, Portugal., Ravache F; ICOM France, 1 Rue Brindejonc des Moulinais, 31500 Toulouse, France., Barandas M; Associação Fraunhofer Portugal Research, Rua Alfredo Allen 455/461, 4200-135 Porto, Portugal.
Jazyk: angličtina
Zdroj: Sensors (Basel, Switzerland) [Sensors (Basel)] 2023 Jul 19; Vol. 23 (14). Date of Electronic Publication: 2023 Jul 19.
DOI: 10.3390/s23146511
Abstrakt: The study of Domain Generalization (DG) has gained considerable momentum in the Machine Learning (ML) field. Human Activity Recognition (HAR) inherently encompasses diverse domains (e.g., users, devices, or datasets), rendering it an ideal testbed for exploring Domain Generalization. Building upon recent work, this paper investigates the application of regularization methods to bridge the generalization gap between traditional models based on handcrafted features and deep neural networks. We apply various regularizers, including sparse training, Mixup, Distributionally Robust Optimization (DRO), and Sharpness-Aware Minimization (SAM), to deep learning models and assess their performance in Out-of-Distribution (OOD) settings across multiple domains using homogenized public datasets. Our results show that Mixup and SAM are the best-performing regularizers. However, they are unable to match the performance of models based on handcrafted features. This suggests that while regularization techniques can improve OOD robustness to some extent, handcrafted features remain superior for domain generalization in HAR tasks.
Databáze: MEDLINE
Nepřihlášeným uživatelům se plný text nezobrazuje