Coursera Flash Sale
40% Off Coursera Plus for 3 Months!
Grab it
Learn about Private Evolution (PE), a groundbreaking training-free framework for generating differentially private synthetic data in this 46-minute Google TechTalk. Discover how PE revolutionizes privacy-preserving machine learning by treating foundation models as blackboxes and utilizing only their inference APIs, eliminating the need for training differentially private generative models. Explore how this innovative approach matches or outperforms state-of-the-art methods in the fidelity-privacy trade-off across both image and text domains while enabling the use of advanced open-source models like Mixtral and API-based models such as GPT-3.5. Understand the computational efficiency advantages of PE compared to existing methods and examine recent extensions including integration with non-neural-network data synthesis tools, fusion of knowledge from multiple models for differentially private data synthesis, and applications in federated learning. Gain insights into how this framework unlocks the full potential of foundation models in privacy-preserving machine learning and accelerates the adoption of differentially private synthetic data across various industries.
