Adaptive Mixtures of Local Experts

Адаптивные смеси локальных экспертов
Geoffrey E. Hinton, Michael I. Jordan, Steven J. Nowlan, Robert A. Jacobs
1991-02-01

adaptive mixtures of local expertscompetitive learningmodular networkssupervised learningvowel discrimination
We present a new supervised learning procedure for systems composed of many separate networks, each of which learns to handle a subset of the complete set of training cases. The new procedure can be viewed either as a modular version of a multilayer supervised network, or as an associative version of competitive learning. It therefore provides a new link between these two apparently different approaches. We demonstrate that the learning procedure divides up a vowel discrimination task into appropriate subtasks, each of which can be solved by a very simple expert network.
1
In a vowel discrimination task, the method partitions the problem into appropriate subtasks handled by simple expert networks.
2
Introduces a supervised learning procedure for systems composed of multiple expert networks, each specializing in a subset of training cases.
3
It establishes a conceptual connection between modular supervised networks and competitive learning approaches.
4
The procedure can be interpreted as either a modular multilayer supervised network or an associative form of competitive learning.

A supervised learning system composed of multiple local expert networks for vowel discrimination

Adaptive decomposition of the training cases into subtasks and coordination of expert networks to solve those subtasks

Publication Details
Publication Date
1991-02-01
Journal
Publisher
ISSN
Cited by
5309
Access Type
Author Information
Authors
Geoffrey E. Hinton
Michael I. Jordan
Steven J. Nowlan
Robert A. Jacobs
Explore further
Open the scid.ai AI chat with a ready-made request: it will find papers on a similar topic and help build a literature review.
Find similar papers in the chat →
Make a presentation
100%