# Ilya Razenshteyn

## GENERAL INFORMATION

I am a researcher in the Machine Learning and Optimization group at Microsoft Research Redmond.

## TRIVIA

During Fall 2017, I was a Simons Junior Fellow at Columbia University hosted by Alexandr Andoni. In August 2017 I graduated with Ph.D. in Computer Science from MIT Computer Science and AI Laboratory, where I was privileged to be advised by Piotr Indyk. Check out my award-winning thesis “High-Dimensional Similarity Search and Sketching: Algorithms and Hardness”. I graduated in June 2012 from Moscow State University with B.S. in Mathematics; my advisors were great Maxim Babenko and Sasha Shen.

## NEWS

• 06/12/19: New paper “Provably Robust Deep Learning via Adversarially Trained Smoothed Classifiers”
• 06/11/19: I have given a talk “Adversarial Examples from Computational Constraints” at ICML 2019
• 05/26/19: I have given a talk “Hölder Homeomorphisms and Approximate Nearest Neighbors” at Workshop on Metric Embeddings and Dimensionality Reduction at Northwestern University
• 05/17/19: I have given a talk “Extension theorems, dimension reduction, and clustering” at Simons Collaboration on Algorithms & Geometry Annual Meeting
• 04/21/19: One paper accepted to ICML 2019
• 04/18/19: One paper accepted to COLT 2019
• 04/03/19: New paper “SANNS: Scaling Up Secure Approximate $k$-Nearest Neighbors Search”
• 02/11/19: One paper accepted to STOC 2019
• 02/07/19: New paper “On Mean Estimation for General Norms with Statistical Queries”
• 01/26/19: New paper “Learning Sublinear-Time Indexing for Nearest Neighbor Search”

## MISCELLANEOUS

My wife does Theory as well!

Program Committees: STOC 2019, SODA 2019, CSR 2018, SEA 2016.

## CODE

• FALCONN: a highly optimized C++ library (with Python bindings) for similarity search based on LSH.
• QuadSketch: a simple and fast algorithm for compressing Euclidean metrics.
• pytorch_kmeans: implementation of the k-means algorithm in PyTorch that works for large datasets.