Scaling laws in wearable human activity recognition
Tom Hoddes, Alex Bijamov, Saket Joshi, Daniel Roggen, Ali Etemad, Robert Harle, David Racz · 2025
Read Scaling laws in wearable human activity recognition by Tom Hoddes, Alex Bijamov, Saket Joshi, Daniel Roggen, Ali Etemad, Robert Harle, David Racz (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2502-03364.txt.
Description
Tom Hoddes, Alex Bijamov, Saket Joshi, Daniel Roggen, Ali Etemad, Robert Harle, and David Racz establish scaling laws for wearable activity recognition, in a 2025 paper. Scaling laws link model size and data volume to performance. They are well known in language and vision. This field, reading human activity from body sensors, lacked them. The team ran an exhaustive grid search. Both pretraining data amount and model size were varied.
Pretraining loss followed a power law in data and size. Adding more users beat adding more data per user. Data diversity, they conclude, drives gains. That finding contrasts with some earlier reports. The laws carried over to real tasks. Three benchmark datasets covered postures, walking styles, and daily activities. The authors suggest some past studies used undersized models. Revisiting them with proper capacity may change conclusions. The laws give builders a principled way to size models and datasets.
Catalog record
| CREATOR | Tom Hoddes, Alex Bijamov, Saket Joshi, Daniel Roggen, Ali Etemad, Robert Harle, David Racz |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 32 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2502-03364