Approximate Nearest Neighbors Methods for Learning and Vision

Dec 13


Presentation Abstracts

Organizers &

Morning Session     (7:30 AM - 10:30 AM)

  7:30 Opening remarks and overview Trevor Darrell
MIT Computer Science and Artificial Intelligence Lab
  7:40 Locality-Sensitive Hashing Scheme Based on p-Stable Distributions Mayur Datar
Stanford University
  8:00 Adaptive Mean Shift Based Clustering in High Dimensions Ilan Shimshoni
  8:20 Approximate Nearest Neighbor Regression in Very High Dimensions Stefan Schaal
University of Southern California
  8:40 Q & A
  8:50 Break
  9:00 Object recognition using locality sensitive hashing of shapemes Jitendra Malik
University of California, Berkley
  9:20 Embedding Earth-Mover Distance into Normed Spaces Piotr Indyk
MIT Computer Science and Artificial Intelligence Lab
  9:40 Moderated discussion
  10:20 'Spotlight' presentations

BREAK: lunch, activities, posters.     (10:30 AM - 4:00 PM)

Learning similarity metrics for vision problems Lihong Li
University of Alberta
Regularized Nearest Neighbor Darrin P. Lewis
Columbia University
Fast Contour Matching Using Approximate Earth Mover's Distance Kristen Grauman

Afternoon Session     (4:00 PM - 7:00 PM)

  4:00 On the Risk of the Approximate Nearest Neighbor Classifier John Fisher
  4:20 Computing K-NN classifications without finding the K-NN Andrew Moore
Carnegie Mellon University
  4:40 Nearest Neighbors in Metric Spaces Ken Clarkson
Bell Labs
  5:00 Moderated discussion
  5:25 Break
  5:30 Fast Example-Based Estimation with Parameter-Sensitive Hashing Greg Shakhnarovich
MIT Computer Science and Artificial Intelligence Lab
  5:50 Limitations on the performance and approximation of high dimensional nearest neighbor searches Jonathan Goldstein
Microsoft Research
  6:10 Approximate Nearest Neighbor Retrieval Using Euclidean Embeddings Vassilis Athitsos
Boston University
  6:30 Moderated discussion
  6:55 Closing remarks Trevor Darrell
MIT Computer Science and Artificial Intelligence Lab

gregory @