The supervised IBP: Neighbourhood preserving infinite latent feature models

Quadrianto, Novi and Sharmanska, Viktoriia and Knowles, David A and Ghahramani, Zoubin (2013) The supervised IBP: Neighbourhood preserving infinite latent feature models. In: UAI: Uncertainty in Artificial Intelligence, July 11-15, 2013, Bellevue, Washington, USA.

[img] Text
QuaShaKnoGha13.pdf - Accepted Version
Available under License All rights reserved.
[IST-2013-137-v1+1]
Download (1090Kb)
Official URL: http://auai.org/uai2013/prints/proceedings.pdf

Abstract

We propose a probabilistic model to infer supervised latent variables in the Hamming space from observed data. Our model allows simultaneous inference of the number of binary latent variables, and their values. The latent variables preserve neighbourhood structure of the data in a sense that objects in the same semantic concept have similar latent values, and objects in di*erent concepts have dissimilar latent values. We formulate the supervised in*nite latent variable problem based on an intuitive principle of pulling objects together if they are of the same type, and pushing them apart if they are not. We then combine this principle with a exible Indian Bu*et Process prior on the latent variables. We show that the inferred supervised latent variables can be directly used to perform a nearest neighbour search for the purpose of retrieval. We introduce a new application of dynamically extending hash codes, and show how to e*ectively couple the structure of the hash codes with continuously growing structure of the neighbourhood preserving in*nite latent feature space.

Item Type: Conference or Workshop Item (Paper)
Subjects: 000 Computer science, knowledge & general works > 000 Computer science, knowledge & systems
000 Computer science, knowledge & general works > 000 Computer science, knowledge & systems > 006 Special computer methods
Research Group: Lampert Group
SWORD Depositor: Sword Import User
Depositing User: Sword Import User
Date Deposited: 12 Aug 2013 08:30
Last Modified: 05 Sep 2017 14:31
URI: https://repository.ist.ac.at/id/eprint/137

Actions (login required)

View Item View Item