Learning with Category-Equivariant Architectures for Human Activity Recognition

TL;DR

This paper introduces CatEquiv, a neural network architecture that encodes categorical symmetries for improved robustness and generalization in human activity recognition from inertial sensor data.

Contribution

The paper presents a novel category-equivariant neural network that systematically encodes symmetries like cyclic shifts and sensor hierarchies for HAR tasks.

Findings

CatEquiv outperforms CNNs under out-of-distribution perturbations.

Enforcing categorical symmetries improves invariance and generalization.

CatEquiv achieves higher robustness without increasing model capacity.

Abstract

We propose CatEquiv, a category-equivariant neural network for Human Activity Recognition (HAR) from inertial sensors that systematically encodes temporal, amplitude, and structural symmetries. We introduce a symmetry category that jointly represents cyclic time shifts, positive gain scalings, and the sensor-hierarchy poset, capturing the categorical symmetry structure of the data. CatEquiv achieves equivariance with respect to the categorical symmetry product. On UCI-HAR under out-of-distribution perturbations, CatEquiv attains markedly higher robustness compared with circularly padded CNNs and plain CNNs. These results demonstrate that enforcing categorical symmetries yields strong invariance and generalization without additional model capacity.