報告人簡介：Tianhao Wang is an assistant professor of computer science at the University of Virginia. His research interests lie in data privacy and security, and their connections to machine learning and cryptography. He obtained his Ph.D. from Purdue University in 2021 and held a postdoc position at Carnegie Mellon University. His work about differentially private synthetic data generation won multiple awards in the NIST competitions.
報告內容介紹：Synthetic data generation serves as a powerful tool for sharing and analyzing data while preserving the privacy of individuals. In this talk, we will explore the cutting-edge techniques in synthetic data generation with a focus on maintaining differential privacy. Our discussion will be divided into two main parts. The first part introduces PRIVIMAGE for private image generation. PRIVIMAGE uses a semantic query function to select pre-training data, enhancing training stability and conserving computational resources. The second part introduces GlucoSynth for private time-series data (glucose trace) generation. The core idea of GlucoSynth is to effectively maintain relationships among glucose events and temporal dynamics.
