Scalable Solution for Approximate Nearest Subspace Search
Abstract
Finding the nearest subspace is a fundamental problem and influential to many applications. In particular, a scalable solution that is fast and accurate for a large problem has a great impact. The existing methods for the problem are, however, useless in a large-scale problem with a large number of subspaces and high dimensionality of the feature space. A cause is that they are designed based on the traditional idea to represent a subspace by a single point. In this paper, we propose a scalable solution for the approximate nearest subspace search (ANSS) problem. Intuitively, the proposed method represents a subspace by multiple points unlike the existing methods. This makes a large-scale ANSS problem tractable. In the experiment with 3036 subspaces in the 1024-dimensional space, we confirmed that the proposed method was 7.3 times faster than the previous state-of-the-art without loss of accuracy.
- Publication:
-
arXiv e-prints
- Pub Date:
- March 2016
- DOI:
- 10.48550/arXiv.1603.08810
- arXiv:
- arXiv:1603.08810
- Bibcode:
- 2016arXiv160308810I
- Keywords:
-
- Computer Science - Computer Vision and Pattern Recognition