Advanced UX with tensorflow.js
summary
A live browser demo uses tensorflow.js to train an RNN (options: GRU or LSTM) on mouse sequences and predict which button will be clicked next, highlighting it before the click; training runs locally in the page using WebGL and the model and data never leave the browser. Results show 70% accuracy on held-out data, built from 369 samples collected by hovering, using RNN/GRU/LSTM in tensorflow.js and a confusion matrix to inspect per-button errors.
AI-written summary · about
chapters (9)
- 0:00Start
- 3:19Preload Resources
- 13:05The Confusion Matrix
- 17:38Probability per Button
- 19:38Tensor Flow Test
- 20:00Tensorflow Namespace
- 23:22Questions
- 25:42Model Definition
- 26:37Training
related talks
13mMake your mouse move like a human - modern AI directly in web browser
25mUsing Machine Learning To Improve UI/UX
26mGettin’ into the (Tensor)Flow
21mMachine Learning & JS Take your PWA to the next level
31mA Feasibility of On-Device ML for Adaptive User Interfaces in Frontend Development
15mMachine learning for next gen web apps with TensorFlow.js
36mUse Ai in your Website using Tensorflow.js
29mSuperpowers for next gen web apps: Machine learning