Learning the Ropes of Synthetic Data
summary
A practical primer on synthetic data that shows how models estimate a probability density, resample to produce avatars, and preserve population-level utility while removing unique records — e.g., a NYT “Barry” lookup hits 30 Barry matches (15 when limited to 1968) illustrating reidentification risk; a brain-scan example where small red regions disappear to protect tiny cohorts; Titanic synthetic outputs match correlations (~0.98–1) and introduce a censored token and three >$500 fares removed to protect outliers.
AI-written summary · about
chapters (7)
- 0:00Introduction to synthetic data
- 2:54Synthetic data basics
- 4:21Utility vs privacy trade-off
- 7:37Generating synthetic data
- 9:05Testing utility and privacy
- 12:36Real-world example Titanic
- 13:17Future of synthetic data
related talks
29mDo I need to be Dr. Frankenstein to create real-ish synthetic data?
21mCrafting Flight Simulations: The Art of Synthetic Data Generation
15mThe tidysynthesis R package
37mHow to Start Critical Thinking in Data Science
16mThe Secret Life of Data - Part 3
47mA Practical(ish) Introduction to Data Science
35mTowards Interpretable Data Science
48mHuman data interactions: experiments in how humans analyze data