Shiny for Python Machine Learning Apps with pandas, scikit-learn and TensorFlow
summary
Demonstrates building interactive ML front ends that wire directly to pandas scikit-learn TensorFlow, using Shiny for Python’s inferred reactivity to only recompute what changes and save heavy GPU/CPU work. Shows concrete apps: image saliency heatmaps and pixel-occlusion tests for vision models, a calibration/fairness explorer, and a hangman front end built in about an hour and used the same day. Deploys models serverless via shiny live and Pyodide (app encoded in a long URL so users run it in-browser with no hosting cost), and scales to production.
AI-written summary · about
chapters (10)
- 0:00Shiny for Python intro
- 1:31Why I use Shiny for Python
- 3:38ML workflows with Shiny
- 5:27Interactivity for deeper insights
- 7:54Interactivity in ML
- 9:43Why reactivity matters
- 12:10Inferred reactivity
- 14:56Scalability to production
- 17:05Serverless with Shiny Live
- 19:37Closing & Q&A
related talks
31mTips to Level-Up Your Shiny for Python Applications
20mSupercharge Your Shiny (for Python) App: Unleashing Interactive Jupyter Widgets
41mPractical Machine Learning for Front-End Developers
48mDeep Learning in the Browser
34mMachine Learning in the browser
22mVue + TensorFlow.js: Building Your First ML Powered App
34mMachine learning for front-end developers
37mSoftware Development for Machine Learning in Python