PhD Student
School of Computer Science
University of Massachusetts Amherst

gbernstein@cs.umass.edu

Resume [PDF]

Introduction

I am a PhD student in the College of Information and Computer Sciences at UMass Amherst. I work with Professor Dan Sheldon in the Machine Learning for Data Science laboratory.

I recieved a B.S. in Applied & Engineering Physics in 2010 and an MEng in Computer Science in 2011, both from Cornell University. From 2011 to 2014 I worked at MIT Lincoln Laboratory, where I developed Machine Learning techniques to aid intelligence analysts. During grad school I've interned at McKesson Relay Health, where I leveraged data science to assist a hospital nurse in prioritizing patients, and twice at Amazon, where I worked to improve product and page recommendations using embeddings and music query correction using neural machine translation.

Research

I like making algorithmic hammers with which other scientists can hit nails in their domain. My focus is on problems requiring the analysis of aggregate population data but for which only noisy observations can be made. This mainly includes developing novel techniques that allow for the adaptation of machine learning algorithms to the framework of Differential Privacy. Private inference means modelers can draw population-level conclusions while at the same time data owners can ensure the individuals in sensitive data sets, e.g. health care or location tracking, remain protected. I have also used the same techniques to better understand continent-wide bird migration.

Publications

UMass Amherst

Differentially Private Bayesian Inference for Exponential Families.[PDF] Garrett Bernstein & Daniel Sheldon
To appear NIPS 2018
Differentially Private Learning of Undirected Graphical Models using Collective Graphical Models.[PDF][Video]
Garrett Bernstein, Ryan McKenna, Tao Sun, Daniel Sheldon, Gerome Miklau, Michael Hay
Proceedings of the 32nd International Conference on Machine Learning (ICML) 2017
Innovative Visualizations Shed Light on Avian Nocturnal Migration.[PDF]
Judy Shamoun-Baranes, Andrew Farnsworth, Garrett Bernstein, et al.
PLOS ONE, 2016
Consistently Estimating Markov Chains with Noisy Aggregate Data.[PDF]
Garrett Bernstein, and Daniel Sheldon
Proceedings of the 19th International Conference on Artificial Intelligence and Statistics (AISTATS 2016)
Inference in a partially observed queueing model with applications in ecology. [PDF]
Kevin Winner, Garrett Bernstein, and Daniel Sheldon
Proceedings of the 32nd International Conference on Machine Learning (ICML) 2015

MIT Lincoln Laboratory

Bayesian Discovery of Threat Networks [PDF]
Steven Smith, Ken Senne, Ed Kao, Garrett Bernstein, and Scott Philips
IEEE Transactions on Signal Processing
Bayesian Network Detection Using Absorbing Markov Chains [PDF]
Steven Smith, Ed Kao, Ken Senne, Garrett Bernstein
ICASSP 2014
Stochastic Agent-Based Simulations of Social Networks [PDF]
Garrett Bernstein and Kyle O'Brien
Proceedings of 46th Annual Simulation Symp., San Diego, 7-10 April 2013
Best Paper Award
Covert Network Detection [PDF]
Steven Smith, Ken Senne, Scott Philips, Ed Kao, Garrett Bernstein
Lincoln Laboratory Journal. Vol. 20(1). 2013 : 47-61