Quadrianto, Novi and Lampert, Christoph (2011) Learning multi-view neighborhood preserving projections. Published in: Getoor, Lise and Scheffer, Tobias, (eds.) Proceedings of the 28 th International Conference on Machine Learning; Washington, USA; 28 June - 2 July 2011. 425-432. Association for Computing Machinery, New York. ISBN 9781450306195

Preview

PDF - Accepted Version Download (405kB) | Preview

Abstract

We address the problem of metric learning for multi-view data, namely the construction of embedding projections from data in different representations into a shared feature space, such that the Euclidean distance in this space provides a meaningful within-view as well as between-view similarity. Our motivation stems from the problem of cross-media retrieval tasks, where the availability of a joint Euclidean distance function is a prerequisite to allow fast, in particular hashing-based, nearest neighbor queries. We formulate an objective function that expresses the intuitive concept that matching samples are mapped closely together in the output space, whereas non-matching samples are pushed apart, no matter in which view they are available. The resulting optimization problem is not convex, but it can be decomposed explicitly into a convex and a concave part, thereby allowing efficient optimization using the convex-concave procedure. Experiments on an image retrieval task show that nearest-neighbor based cross-view retrieval is indeed possible, and the proposed technique improves the retrieval accuracy over baseline techniques.

Item Type:	Conference Proceedings
Schools and Departments:	School of Engineering and Informatics > Informatics
Subjects:	Q Science > Q Science (General)
Depositing User:	Novi Quadrianto
Date Deposited:	24 Feb 2014 14:16
Last Modified:	16 Jun 2017 15:02
URI:	http://srodev.sussex.ac.uk/id/eprint/47612

View download statistics for this item

📧 Request an update