BARC talk by Matthew Regehr
Tuesday, April 8, 2025, Matthew Regehr, PhD student at the University of Waterloo, Canada, will give a talk on "A Bias-Accuracy-Privacy Trilemma for Statistical Estimation".
Abstract:
Differential privacy (DP) is a rigorous notion of data privacy, used for private statistics. The canonical algorithm for differentially private mean estimation is to first clip the samples to a bounded range and then add noise to their empirical mean. Clipping controls the sensitivity and, hence, the variance of the noise that we add for privacy. But clipping also introduces statistical bias. This tradeoff is inherent: we prove that no algorithm can simultaneously have low bias, low error, and low privacy loss for arbitrary distributions.
Additionally, we show that under strong notions of DP (i.e., pure or concentrated DP), unbiased mean estimation is impossible, even if we assume that the data is sampled from a Gaussian. On the positive side, we show that unbiased mean estimation is possible under a more permissive notion of differential privacy (approximate DP) if we assume that the distribution is symmetric.
Bio:
Matthew Regehr is a PhD student in the Cheriton School of Computer Science at the University of Waterloo, supervised by Gautam Kamath. Matt's research interests are broadly machine learning theory and differential privacy. Current research projects focus on the theory of privacy accounting, sample-efficient algorithms for locally private hypothesis selection, and fairness in machine learning.
Host:
Rasmus Pagh