Online estimation of a smooth regression function

R. Khasminskii*, R. Liptser

*Corresponding author for this work

Research output: Contribution to journalArticlepeer-review

1 Scopus citations

Abstract

The tracking (recursive) type estimator for a one-dimensional regression estimation problem with equidistant design is proposed. It is proven that, out of the inevitable initial layer, this estimator has the optimal rate of convergence of quadratic risk to zero if the sample size goes to infinity.

Original languageEnglish
Pages (from-to)541-550
Number of pages10
JournalTheory of Probability and its Applications
Volume47
Issue number3
DOIs
StatePublished - 2003

Keywords

  • Equidistant design
  • Nonparametric estimation
  • Online tracking estimator
  • Regression

Fingerprint

Dive into the research topics of 'Online estimation of a smooth regression function'. Together they form a unique fingerprint.

Cite this