ML training on the web: Building Simple ML for Google Sheets
summary
Claims that decision-forest models train in seconds and predict in microseconds using on-device ML via WebAssembly and npm packages. Describes YDF (open source) that shares C++/Python/JS code, supports CSV and object inputs, trains in-browser on CPU, and offers a separate ydf-inference engine. Shows Simple ML for Sheets — a free Google Sheets add-on with 150K installations that uses Google Apps Script to spin up an iframe, train models locally, store models on Google Drive, and return predictions and confidences back into the sheet.
AI-written summary · about
chapters (11)
- 0:00Introduction to web ML
- 1:36Advantages of browser ML
- 2:09Why train on the web
- 3:52Introducing YDF library
- 5:16YDF in Python and JS
- 7:02YDF inference and features
- 7:39Simple ML for Sheets demo
- 8:37Using Simple ML in sheets
- 10:41How simple ML works
- 11:25Technical implementation
- 12:29Future and resources
related talks
19mML at Client Side
46mML on the client: training on squirrels
41mPractical Machine Learning for Front-End Developers
24mMachine learning in the browser
23mNode.js = Machine Learning
24mThe next level apps for Web developers Machine learning on browsers 🤯
39mWebAssembly-Powered Machine Learning
22mBoring but Effective: WebAI That Ships