Visual diagnostics for more informed machine learning
summary
Argument: adding visual diagnostics at each pipeline stage makes machine learning more informed, not just prettier. She demonstrates concrete tools: scikitlearn pandas seaborn Matplotlib, feature plots like radviz and parallel coordinates, pairwise hyperparameter heat maps and validation curve visualizations. A working library called Yellowbrick is on PyPI (0.3 alpha, pip installable) that implements separability tests, class balance reports, prediction error and residual plots, and interactive plans; it exists, runs, and is soliciting testers and contributors.
AI-written summary · about
chapters (6)
- 0:00Introduction
- 7:48The limitations of quick machine learning
- 12:22Visual diagnostics
- 26:19Yellow Brick project
- 30:29Get involved
- 32:58Q&A
related talks
35mVisual Diagnostics for More Effective Machine Learning
38mBuilding a Visualizer in Yellowbrick
39mYellowbrick: Steering Machine Learning with Visual Transformers
43mVisualising Machine Learning Models
01h 33mUnderstanding and diagnosing your machine-learning models
22mVisualizing machine learning algorithms in the Jupyter notebook
30mWords in Space
02h 23mEffective Data Visualization