We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

stat.ME

Change to browse by:

References & Citations

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Statistics > Methodology

Title: Sparse principal component analysis via axis-aligned random projections

Abstract: We introduce a new method for sparse principal component analysis, based on the aggregation of eigenvector information from carefully-selected axis-aligned random projections of the sample covariance matrix. Unlike most alternative approaches, our algorithm is non-iterative, so is not vulnerable to a bad choice of initialisation. We provide theoretical guarantees under which our principal subspace estimator can attain the minimax optimal rate of convergence in polynomial time. In addition, our theory provides a more refined understanding of the statistical and computational trade-off in the problem of sparse principal component estimation, revealing a subtle interplay between the effective sample size and the number of random projections that are required to achieve the minimax optimal rate. Numerical studies provide further insight into the procedure and confirm its highly competitive finite-sample performance.
Comments: 32 pages
Subjects: Methodology (stat.ME); Statistics Theory (math.ST); Machine Learning (stat.ML)
MSC classes: 62H25
Cite as: arXiv:1712.05630 [stat.ME]
  (or arXiv:1712.05630v4 [stat.ME] for this version)

Submission history

From: Richard Samworth [view email]
[v1] Fri, 15 Dec 2017 11:55:39 GMT (286kb,D)
[v2] Sat, 13 Jan 2018 17:35:54 GMT (525kb,D)
[v3] Tue, 24 Jul 2018 10:12:09 GMT (600kb,D)
[v4] Mon, 6 May 2019 16:18:07 GMT (287kb,D)

Link back to: arXiv, form interface, contact.