Publication detail

The Behavior of LWPR Method on Approximation of Multivariable Functions

JANDORA, R.

English title

The Behavior of LWPR Method on Approximation of Multivariable Functions

Type

Paper in proceedings (conference paper)

Language

en

Original abstract

Locally Weighted Projection Regression is an incremental approximation method used in real-time control. It was developed to approximate functions with high-dimensional inputs. When used for problems with multivariable outputs, this algorithm has serious problems when trying to approximate such function. This is caused by simplifications used when creating the incremental version of the statistical method for the local linear model. These simplifications limit the usage of the LWPR method to functions with single-variable output.

Keywords in English

LWPR; locally weighted learning

Released

2005-05-26

Publisher

Ing. Zdeněk Novotný, CSc., Ondráčkova 105, Brno

Location

Brno

ISBN

80-214-2942-9

Book

Honeywell EMI 2005 Proceedings of the International Interdisciplinary Student Competition and Conference

Pages from–to

105–109

Pages count

5

BIBTEX


@inproceedings{BUT17975,
  author="Radek {Jandora}",
  title="The Behavior of LWPR Method on Approximation of Multivariable Functions",
  booktitle="Honeywell EMI 2005 Proceedings of the International Interdisciplinary Student Competition and Conference",
  year="2005",
  number="1.",
  pages="105--109",
  publisher="Ing. Zdeněk Novotný, CSc., Ondráčkova 105, Brno",
  address="Brno",
  isbn="80-214-2942-9"
}