Summary of On-device Online Learning and Semantic Management Of Tinyml Systems, by Haoyu Ren et al.
On-device Online Learning and Semantic Management of TinyML Systems
by Haoyu Ren, Xue Li, Darko Anicic, Thomas A. Runkler
First submitted to arxiv on: 13 May 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI); Databases (cs.DB); Distributed, Parallel, and Cluster Computing (cs.DC)
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 proposed study aims to bridge the gap between prototyping single Tiny Machine Learning (TinyML) models and developing reliable TinyML systems in production. It tackles three main challenges: (1) online learning for adapting local models towards changing field conditions; (2) federated meta-learning for enhancing model generalization across heterogeneous devices with scarce labeled data; and (3) semantic management for joint management of models and devices at scale. The study demonstrates its methods through a basic regression example and assesses them in three real-world TinyML applications: handwritten character image classification, keyword audio classification, and smart building presence detection. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Tiny Machine Learning (TinyML) is making it possible to run machine learning on low-footprint embedded devices. This can be useful for things like recognizing handwritten characters or detecting when someone is at home. The problem is that most current solutions just focus on doing one thing really well, but don’t work well in real-world situations where things are always changing. The study proposes three new ideas to make TinyML more practical: online learning to adapt models to changing conditions, federated meta-learning to share knowledge across devices, and semantic management to organize all the different devices and models at scale. |
Keywords
» Artificial intelligence » Classification » Generalization » Image classification » Machine learning » Meta learning » Online learning » Regression
