Optimal Rates of Aggregation

Tsybakov, Alexandre B.

doi:10.1007/978-3-540-45167-9_23

Optimal Rates of Aggregation

Alexandre B. Tsybakov

COLT 2003 pp. 303-313

doi:10.1007/978-3-540-45167-9_23 /colt/2003/tsybakov2003colt-optimal/

Abstract

We study the problem of aggregation of M arbitrary estimators of a regression function with respect to the mean squared risk. Three main types of aggregation are considered: model selection, convex and linear aggregation. We define the notion of optimal rate of aggregation in an abstract context and prove lower bounds valid for any method of aggregation. We then construct procedures that attain these bounds, thus establishing optimal rates of linear, convex and model selection type aggregation.

PDF COLT Semantic Scholar

Cite

Text

Tsybakov. "Optimal Rates of Aggregation." Annual Conference on Computational Learning Theory, 2003. doi:10.1007/978-3-540-45167-9_23

Markdown

[Tsybakov. "Optimal Rates of Aggregation." Annual Conference on Computational Learning Theory, 2003.](https://mlanthology.org/colt/2003/tsybakov2003colt-optimal/) doi:10.1007/978-3-540-45167-9_23

BibTeX

@inproceedings{tsybakov2003colt-optimal,
  title     = {{Optimal Rates of Aggregation}},
  author    = {Tsybakov, Alexandre B.},
  booktitle = {Annual Conference on Computational Learning Theory},
  year      = {2003},
  pages     = {303-313},
  doi       = {10.1007/978-3-540-45167-9_23},
  url       = {https://mlanthology.org/colt/2003/tsybakov2003colt-optimal/}
}