Learning Multi-View Neighborhood Preserving Projections
Novi Quadrianto, Christoph H. Lampert · 2011
We address the problem of metric learning for multi-view data, namely the construction of embedding projections from data in dif-ferent 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 moti-vation stems from the problem of cross-media retrieval tasks, where the availability of a joint Euclidean distance function is a pre-requisite to allow fast, in particular hashing-based, nearest neighbor queries. We formulate an objective function that ex-presses 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 optimiza-tion problem is not convex, but it can be decomposed explicitly into a convex and a concave part, thereby allowing efficient op-timization using the convex-concave proce-dure. Experiments on an image retrieval task show that nearest-neighbor based cross-view retrieval is indeed possible, and the pro-posed technique improves the retrieval accu-racy over baseline techniques. 1.