Machine Learning and JavaScript. Unleashing the power of sentiment analysis
summary
Shows how to convert Python models to run in-browser using tensorflow.js + WASM, trains a sentiment model on the IMDB dataset with a bi-directional LSTM and reports about 84% accuracy on test data while warning about overfitting, demonstrates a Chrome-side pipeline that fetches Instagram comments, maps words to indices, runs the model and highlights sentiment in-page, notes Pyodide can incur ~9x slowdown and that runtime speed ultimately depends on the client GPU and WebGL/WebGPU support.
AI-written summary · about
chapters (19)
- 0:00Start
- 3:30Typescript
- 5:58What Is Machine Learning
- 6:20Reinforcement Learning
- 6:34Supervised Learning
- 7:37How Machine Learning Actually Works
- 8:05How Machine Learning Works under the Hood
- 10:44Demo
- 13:30Web of Things
- 16:41Porting from Python
- 18:24Tensorflow Js Converter
- 19:36Convolutional Neural Networks
- 27:59Lstm Training Demo
- 29:27Model Overfitting
- 29:56Positive Sentiment Recognition
- 34:53Possible Pitfalls
- 44:03Is There a Fight between Javascript and Python on Machine Learning Arena
- 58:55What Is Inspiring You in Machine Learning
- 1:05:42Do You Use Anything from Your Work in Your Real Life
related talks
39mMachine Learning Magic for Your JavaScript Application
15mThe future of machine learning using JavaScript
57mMachine Learning with JavaScript
16mJavaScript Text Analytics for User Behavior Intelligence
30mBuilding Machine Learning Models on Edge with React
22mKeynote talk - The future of machine learning & JS
44mArtificial Intelligence and Machine Learning in the browser
37mMachine Learning Principles, with JavaScript?