Gavin Brown

About Me

Hello! I am a postdoc at the University of Washington with Sewoong Oh. I completed my PhD at Boston University, where I was advised by Adam Smith.

I work on machine learning and data privacy. The outputs of data analysis depend on the details of individual data points, sometimes heavily. When is this necessary, and when can we avoid it? I am interested in understanding when and why machine learning models memorize large amounts of training examples.

I also study this topic through the lens of differential privacy, a formal framework for reasoning about privacy in data analysis. Here, my main work is in designing algorithms for fundamental statistical problems.

For the first two years of my PhD, I was advised by Peter Chin and worked on applications of machine learning and compressed sensing. Before that, I received a BS in Mathematics from Case Western Reserve University in 2015. My Senior Capstone project was advised by David Gurarie.

Awards

• My work on private mean estimation with Sam Hopkins and Adam Smith shared the COLT 2023 award for Best Student Paper.
• Boston University's Department of Computer Science Awarded me the 2022/23 Research Excellence Award.
• I deeply enjoy teaching, both in and out of the classroom. In 2020, I received a Teaching Fellow Excellence Award from BU's Computer Science Department.

Recent Publications

• Insufficient Statistics Perturbation: Stable Estimators for Private Least Squares.
  Gavin Brown, Jonathan Hayase, Samuel Hopkins, Weihao Kong, Xiyang Liu, Sewoong Oh, Juan C. Perdomo, and Adam Smith.
  COLT 2024.

• Private Gradient Descent for Linear Regression: Tighter Error Bounds and Instance-Specific Uncertainty Estimation.
  Gavin Brown, Krishnamurthy Dvijotham, Georgina Evans, Daogao Liu, Adam Smith, and Abhradeep Thakurta.
  ICML 2024.

• Metalearning with Very Few Samples Per Task.
  Maryam Aliakbarpour, Konstantina Bairaktari, Gavin Brown, Adam Smith, Nathan Srebro, and Jonathan Ullman.
  COLT 2024. Proceedings version.

• Fast, Sample-Efficient, Affine-Invariant Private Mean and Covariance Estimation for Subgaussian Distributions.
  Gavin Brown, Samuel B. Hopkins, and Adam Smith.
  COLT 2023, Best Student Paper.

• Strong Memory Lower Bounds for Learning Natural Models.
  Gavin Brown, Mark Bun, and Adam Smith.
  COLT 2022. Proceedings version.

• Performative Prediction in a Stateful World.
  Gavin Brown, Iden Kalemaj, and Shlomi Hod.
  AISTATS 2022. Proceedings version.
  A preliminary version of this paper appeared at the NeurIPS Workshop on Consequential Decision Making in Dynamic Environments, 2020.

• Covariance-Aware Private Mean Estimation Without Private Covariance Estimation
  Gavin Brown, Marco Gaboardi, Adam Smith, Jonathan Ullman, and Lydia Zakynthinou.
  NeurIPS 2021, Spotlight Presentation. Proceedings version.

• When Is Memorization of Irrevelant Training Data Necessary for High-Accuracy Learning?
  Gavin Brown, Mark Bun, Vitaly Feldman, Adam Smith, and Kunal Talwar.
  STOC 2021. Proceedings version.