Breaking the Privacy-Utility Tradeoff with Deming Regression
summary
Presents using Deming regression to train salary-prediction models on anonymized (privacy-protected) survey data so the model’s quality approaches that of one trained on raw private data. Demonstrates the approach with live coding in a Jupyter notebook, showing how to apply Deming regression in practice to balance privacy and utility.
AI-written summary · about
related talks
27mYou've got trust issues, we've got solutions: Differential Privacy
30mPrivate Generative Models: Balancing Privacy and Utility in AI
14mEstablishing Privacy Metrics for Genomic Data Analysis
33mPractical Data Privacy
38mPrivacy: the Last Stand for Fair Algorithms
26mDoes using data mean giving up privacy?
46mLearning the Privacy-Utility Trade-off with Bayesian Optimization
33mMachine Learning in the age of increasing Data Privacy Consciousness