Build trustworthy ML detection and monitor bias
summary
Shows how Amazon SageMaker Clarify runs a Clarify processing job (no extra cost) to detect pre/post/deployed bias on a census UCI income classifier (predicts >$50k/year) and flags a 2:1 male:female skew and a 6:1 skew among positive labels; trains XGBoost on S3 data, records SHAP-style feature importance where capital gains tops the list, and wires continuous production checks into SageMaker model monitor with alerts to CloudWatch. Includes PDF/HTML reports, notebook artifacts and SageMaker Studio integration.
AI-written summary · about
chapters (8)
- 0:00Introduction
- 0:33Agenda
- 1:30Background in Bias
- 8:29Explainability
- 11:16Why
- 16:04Amazon SageMaker Clarify
- 18:39Demo
- 30:17Recap
related talks
24mUnderstand ML model predictions & biases with Amazon SageMaker Clarify
18mUnderstand your ML models better with greater visibility and transparency
44mImprove explainability of ML models to meet regulatory requirements
41mMachine Learning Model Fairness and Interpretability with AWS
55mEthical Machine Learning
37mStreamlining Machine Learning Operations
33mBuild, train, and deploy ML models using Amazon SageMaker
33mMachine Learning Engineering Done Right