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

Download:

Current browse context:

stat.ML

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 > Machine Learning

Title: Linear Convergence of the Subspace Constrained Mean Shift Algorithm: From Euclidean to Directional Data

Abstract: This paper studies linear convergence of the subspace constrained mean shift (SCMS) algorithm, a well-known algorithm for identifying a density ridge defined by a kernel density estimator. By arguing that the SCMS algorithm is a special variant of a subspace constrained gradient ascent (SCGA) algorithm with an adaptive step size, we derive linear convergence of such SCGA algorithm. While the existing research focuses mainly on density ridges in the Euclidean space, we generalize density ridges and the SCMS algorithm to directional data. In particular, we establish the stability theorem of density ridges with directional data and prove the linear convergence of our proposed directional SCMS algorithm.
Comments: 75 pages, 9 figures
Subjects: Machine Learning (stat.ML); Statistics Theory (math.ST); Methodology (stat.ME)
MSC classes: 62G05, 49Q12, 62H11
Cite as: arXiv:2104.14977 [stat.ML]
  (or arXiv:2104.14977v1 [stat.ML] for this version)

Submission history

From: Yikun Zhang [view email]
[v1] Thu, 29 Apr 2021 01:46:35 GMT (9367kb,D)

Link back to: arXiv, form interface, contact.