Abstract
The goal of subspace learning is to find a k-dimensional subspace of ℝd, such that the expected squared distance between instance vectors and the subspace is as small as possible. In this paper we study subspace learning in a partial information setting, in which the learner can only observe r ≤ d attributes from each instance vector. We propose several efficient algorithms for this task, and analyze their sample complexity.
| Original language | English |
|---|---|
| Article number | 52 |
| Number of pages | 21 |
| Journal | Journal of Machine Learning Research |
| Volume | 17 |
| State | Published - 1 Apr 2016 |
Bibliographical note
Funding Information:We thank the anonymous reviewers for their helpful comments. We also thank Amit Daniely for helpful discussions. This research has been supported by ISF no 1673/14.
Publisher Copyright:
©2016 Alon Gonen, Dan Rosenbaum, Yonina C. Eldar and Shai Shalev-Shwartz.
Keywords
- Budgeted learning
- Learning theory
- Learning with partial information
- Principal components analysis
- Statistical learning