On Data (In-)Dependent Hashing
Add to your list(s)
Download to your calendar using vCal
If you have a question about this talk, please contact Konstantina Palla.
I will provide an overview of techniques to perform approximate nearest neighbor (ANN) search in massive datasets.
The ANN search has wide-ranging applications, among others, in information retrieval for finding near-duplicate pages,
in computer graphics for completing scenes, and in collaborative filtering. The most widely used approach that is particularly
suitable for high-dimensional data is to build similarity-preserving hash functions which map similar data points to nearby codes.
These hashing methods can be sub-divided into two main categories: data independent and data dependent methods.
I will cover the locality-sensitive hashing (LSH)-based methods as a representative of the data independent approach.
I will show how to build LSH that preserves hamming distance, cosine similarity, and Jaccard index. I will briefly mention some
of recent machine learning based data dependent approaches such as spectral hashing and other loss-based hashing.
To make things a bit closer to home research, I will also try to show some potentials of hashing for Gaussian Process Regression.
This talk is part of the Machine Learning Reading Group @ CUED series.
This talk is included in these lists:
Note that ex-directory lists are not shown.
|