Differentially Private Synthetic Data without Training

Name: Differentially Private Synthetic Data without Training
Uploaded: 2026-01-27T00:00:00.000Z
Duration: 45 min 44 s
Description: Microsoft Research's Zinan Lin presents Private Evolution (PE), a way to generate differentially private (DP) synthetic data without training. It only calls a model's inference APIs: one generates random samples and o...

Google TechTalks · 45:44 · Watch on YouTube

Microsoft Research's Zinan Lin presents Private Evolution (PE), a way to generate differentially private (DP) synthetic data without training. It only calls a model's inference APIs: one generates random samples and o...

Read the full summary on retube

한국어