Summary of Balancing Continual Learning and Fine-tuning For Human Activity Recognition, by Chi Ian Tang et al.
Balancing Continual Learning and Fine-tuning for Human Activity Recognition
by Chi Ian Tang, Lorena Qendro, Dimitris Spathis, Fahim Kawsar, Akhil Mathur, Cecilia Mascolo
First submitted to arxiv on: 4 Jan 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Signal Processing (eess.SP)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary The paper explores the adoption and adaptation of two continual self-supervised learning models, CaSSLe and Kaizen, for wearable-based Human Activity Recognition (HAR). The authors aim to develop HAR systems that are tailored to users’ needs through continual learning. They propose a semi-supervised approach called Kaizen, which balances representation learning and down-stream classification using labelled and unlabelled data. The paper also investigates the importance of different loss terms in continual learning and evaluates the trade-off between knowledge retention and learning from new tasks. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Wearable-based Human Activity Recognition (HAR) is a way to understand how people behave. To make HAR better, we need to make it learn and adapt as people do new things. Right now, there are limited ways to collect data for this kind of learning. The paper looks at two new methods called CaSSLe and Kaizen that can help us with this problem. They use special types of learning called self-supervised and semi-supervised learning to make HAR systems better. |
Keywords
* Artificial intelligence * Activity recognition * Classification * Continual learning * Representation learning * Self supervised * Semi supervised
