Beyond Bias: Crafting Fairness in Pythonic Machine Learning Systems
summary
Only 13% of ML models make it to production and here the workflow shows concrete fixes: open German credit data, apply SMOTE for class balance, train an XGBoost classifier that reaches about 80% accuracy on a small dataset, then evaluate statistical parity difference and predictive parity to check group gaps (reported as small). Explanations come from LIME highlighting contributive words, and testing combines unit-style assertions with stochastic model metrics to catch regressions.
AI-written summary · about
chapters (15)
- 0:00Introduction to talk
- 1:08Machine learning basics
- 2:16Blackbox problem examples
- 5:26Reasons for biased models
- 9:18Building fair models
- 10:43Loan classification demo
- 14:33Fairness techniques
- 17:24Why test ML systems
- 18:52Sentiment analysis testing
- 21:41Responsibility for model errors
- 23:51Trustworthy systems need
- 25:34Importance of explainability
- 27:01Technical debt in ML
- 29:06Mitigation strategies
- 30:15Q&A session
related talks
01h 30mFairness Tales: How To Measure And Mitigate Unfair Bias in Machine Learning Models
42mFairness in Machine Learning: From Theory to Practice
46mA look at the methods to detect and try to remove bias in ML models
37mKeynote: Ethical Machine Learning: Creating Fair Models in an Unjust World
36mMachine Learning Fairness: Lessons Learned
47mEthical Machine Learning
31mFlagging *and* Fixing Bias in ML
23mRemoving Unfair Bias in the Machine Learning Pipeline