Understand your ML models better with greater visibility and transparency
summary
Hands-on demo of Amazon SageMaker Clarify shows concrete outputs: pre/post-training bias reports (JSON, HTML, PDF) from a credit dataset with age > 40 as a sensitive group, feature attributions computed with kernel SHAP and plotted via the SHAP library, and continuous checks via SageMaker Model Monitor tied into Data Wrangler and Pipelines. Customers report using it for credit scoring and a 16‑factor goal‑probability model; Clarify is generally available, billed by processing time, and works across SageMaker regions.
AI-written summary · about
chapters (12)
- 0:00Intro
- 0:27Agenda
- 0:45ML Development
- 2:00Bias
- 3:29Existing tools
- 4:22Clarify
- 4:43Clarify integration
- 6:45Bundesliga
- 7:41Demo
- 16:03Recap
- 16:29Resources
- 17:25Outro
related talks
24mUnderstand ML model predictions & biases with Amazon SageMaker Clarify
44mImprove explainability of ML models to meet regulatory requirements
33mBuild trustworthy ML detection and monitor bias
41mMachine Learning Model Fairness and Interpretability with AWS
37mStreamlining Machine Learning Operations
33mBuild, train, and deploy ML models using Amazon SageMaker
55mEthical Machine Learning
52mGenerate ML predictions without writing any code