Unlearning in feed-forward multi-nets

L Spaanenburg

Unlearning in feed-forward multi-nets

L Spaanenburg^*

^*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › Academic › peer-review

Abstract

Multi-nets promise an improved performance over monolithic neural networks by virtue of their distributed implementation. Modular neural networks are multi-nets based on an judicious assembly of functionally different parts. This can be viewed as again a monolithic network, but with more complex neurons (the neural modules). Therefore they will share the same learning problems, notably the unlearning effect. In this paper we will look more closely into the reasons for unlearning and discuss how this can be applied to detect novelties.

Original language	English
Title of host publication	ARTIFICIAL NEURAL NETS AND GENETIC ALGORITHMS
Editors	Kurkova, NC Steele, R Neruda, M Karny
Place of Publication	VIENNA
Publisher	Springer
Pages	106-109
Number of pages	4
ISBN (Print)	3-211-83651-9
Publication status	Published - 2001
Event	International Conference on Artificial Neural Networks and Genetic Algorithms (ICANNGA) - , Czech Republic Duration: 22-Apr-2001 → 25-Apr-2001

Publication series

Name	SPRINGER COMPUTER SCIENCE
Publisher	SPRINGER-VERLAG WIEN

Other

Other	International Conference on Artificial Neural Networks and Genetic Algorithms (ICANNGA)
Country/Territory	Czech Republic
Period	22/04/2001 → 25/04/2001

Keywords

OPTIMIZATION

Handle.net

Persistent link

Cite this

@inproceedings{e2fd48cb96084c52bc5710ce90662a97,

title = "Unlearning in feed-forward multi-nets",

abstract = "Multi-nets promise an improved performance over monolithic neural networks by virtue of their distributed implementation. Modular neural networks are multi-nets based on an judicious assembly of functionally different parts. This can be viewed as again a monolithic network, but with more complex neurons (the neural modules). Therefore they will share the same learning problems, notably the unlearning effect. In this paper we will look more closely into the reasons for unlearning and discuss how this can be applied to detect novelties.",

keywords = "OPTIMIZATION",

author = "L Spaanenburg",

year = "2001",

language = "English",

isbn = "3-211-83651-9",

series = "SPRINGER COMPUTER SCIENCE",

publisher = "Springer",

pages = "106--109",

editor = "Kurkova and NC Steele and R Neruda and M Karny",

booktitle = "ARTIFICIAL NEURAL NETS AND GENETIC ALGORITHMS",

note = "International Conference on Artificial Neural Networks and Genetic Algorithms (ICANNGA) ; Conference date: 22-04-2001 Through 25-04-2001",

}

TY - GEN

T1 - Unlearning in feed-forward multi-nets

AU - Spaanenburg, L

PY - 2001

Y1 - 2001

N2 - Multi-nets promise an improved performance over monolithic neural networks by virtue of their distributed implementation. Modular neural networks are multi-nets based on an judicious assembly of functionally different parts. This can be viewed as again a monolithic network, but with more complex neurons (the neural modules). Therefore they will share the same learning problems, notably the unlearning effect. In this paper we will look more closely into the reasons for unlearning and discuss how this can be applied to detect novelties.

AB - Multi-nets promise an improved performance over monolithic neural networks by virtue of their distributed implementation. Modular neural networks are multi-nets based on an judicious assembly of functionally different parts. This can be viewed as again a monolithic network, but with more complex neurons (the neural modules). Therefore they will share the same learning problems, notably the unlearning effect. In this paper we will look more closely into the reasons for unlearning and discuss how this can be applied to detect novelties.

KW - OPTIMIZATION

M3 - Conference contribution

SN - 3-211-83651-9

T3 - SPRINGER COMPUTER SCIENCE

SP - 106

EP - 109

BT - ARTIFICIAL NEURAL NETS AND GENETIC ALGORITHMS

A2 - Kurkova, null

A2 - Steele, NC

A2 - Neruda, R

A2 - Karny, M

PB - Springer

CY - VIENNA

T2 - International Conference on Artificial Neural Networks and Genetic Algorithms (ICANNGA)

Y2 - 22 April 2001 through 25 April 2001

ER -

Unlearning in feed-forward multi-nets

Abstract

Publication series

Other

Keywords

Handle.net

Fingerprint

Cite this