Reproduce “Transfer Learning in Human Activity Recognition: A Survey” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Sourish Gunesh Dhekane, Thomas Ploetz
A comprehensive survey from Georgia Institute of Technology reviews 205 papers on transfer learning in sensor-based Human Activity Recognition, categorizing approaches that address data scarcity and domain shift in smart home and wearable applications. The analysis identifies prevalent deep learning techniques and outlines future research directions for building generalizable HAR systems.
Key insights
- Transfer learning effectively addresses data scarcity and domain shift, enhancing data efficiency, generalization, and adaptability in HAR systems.
- Deep learning approaches, including pretrain-then-finetune, adversarial learning, and self-supervised learning, are increasingly prevalent and successful in HAR transfer learning.
- Challenges like multi-faceted domain shifts, privacy-preserving distributed learning, and adaptation to dynamic environments remain key areas for future research.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.