Labeling against all odds
summary
Covers labeling challenges at PerimeterX including learning from partially labeled data, handling datasets with potential mislabels, and building feedback loops on model predictions. Reviews academic approaches for these problems and presents PerimeterX case studies illustrating different practical solutions.
AI-written summary · about
chapters (14)
- 0:00Intro
- 0:49YOU ARE A DATA SCIENTIST IN THE WEB SECURITY WORLD
- 3:23Unsupervised Learning Model
- 5:36From Heuristics to Labels
- 8:32Manual Labeling
- 9:39Active Learning
- 12:39Automatic Labeling Functions
- 16:55Noisy Labels
- 19:57Modeling? How?
- 21:06Semi Supervised Learning
- 22:09Self-Training
- 23:42Co-Training
- 26:14Labeling Against All Odds
- 27:46What Did We Learn?
