Differential Analysis
Differential analysis encompasses a broad range of techniques aiming to analyze and manipulate data while addressing challenges like privacy, efficiency, and robustness. Current research focuses on improving the efficiency and accuracy of differential equation-based models for tasks such as image restoration and generative modeling, as well as developing novel algorithms for differentially private machine learning and optimizing large-scale problems through enhanced grouping methods. These advancements are significant for improving the performance and privacy guarantees of machine learning models, enabling more efficient solutions for complex engineering problems, and facilitating reliable data analysis in sensitive applications.
Papers
The Last Iterate Advantage: Empirical Auditing and Principled Heuristic Analysis of Differentially Private SGD
Thomas Steinke, Milad Nasr, Arun Ganesh, Borja Balle, Christopher A. Choquette-Choo, Matthew Jagielski, Jamie Hayes, Abhradeep Guha Thakurta, Adam Smith, Andreas Terzis
Private and Communication-Efficient Federated Learning based on Differentially Private Sketches
Meifan Zhang, Zhanhong Xie, Lihua Yin