Best-First Model Merging for Dynamic Learning and Recognition

TitleBest-First Model Merging for Dynamic Learning and Recognition
Publication TypeTechnical Report
Year of Publication1992
AuthorsOmohundro, S.
Other Numbers709
Abstract

"Best-first model merging" is a general technique for dynamically choosing the structure of a neural or related architecture while avoiding overfitting. It is applicable to both learning and recognition tasks and often generalizes significantly better than fixed structures. We demonstrate the approach applied to the tasks of choosing radial basis functions for function learning, choosing local affine models for curve and constraint surface modeling, and choosing the structure of a balltree or bumptree to maximize efficiency of access.

URLhttp://www.icsi.berkeley.edu/ftp/global/pub/techreports/1992/tr-92-004.pdf
Bibliographic Notes

ICSI Technical Report TR-92-004

Abbreviated Authors

S. M. Omohundro

ICSI Publication Type

Technical Report