The Geometry of Similarity Search

Lecture

Speaker

  • Portrait photo of Alexandr Andoni
    Alexandr Andoni Principal Investigator, Simons Collaboration on Algorithms and Geometry

When

  • – ET

Where

Gerald D. Fischbach Auditorium

160 5th Avenue

New York, NY 10010, United States

Open Map for Gerald D. Fischbach Auditorium

How does one efficiently search for similar items in a large dataset, say, of images? This is a classic algorithmic task that arises when dealing with massive datasets. Its applications span many areas, including information retrieval, machine learning, data mining, computer vision, signal processing, bioinformatics, etc. For example, this task underlies the classic ‘nearest neighbor’ classification rule in machine learning. The natural solution for similarity search — to scan the entire dataset, comparing a given item to each item in the dataset — is prohibitively slow for modern datasets.

Alexandr Andoni will describe how efficient solutions for similarity search benefit from the tools and perspectives of high-dimensional geometry. The latter emerged as the natural language of similarity search: e.g., a 20 x 20 pixel image naturally corresponds to a 400-dimensional vector, one coordinate per pixel. Andoni will show how geometric tools such as dimension reduction and randomized space partitions enable efficient similarity search algorithms.

More Events

Initiative for Computational Catalysis

Machine Learning for Computational Catalysis

Workshop

November 16–18, 2026

Speaker

François Gallaire, Ph.D.

Science, Society & Culture

Statistics in the Era of Computation

Presidential Lecture

November 18, 2026

6–7 p.m. ET

Speaker

Kyunghyun Cho