Translating feed-forward nets to SOM-like maps

B J vanderZwaag; Lambert Spaanenburg; C Slump

Translating feed-forward nets to SOM-like maps

B J vanderZwaag, Lambert Spaanenburg, C Slump

Department of Electrical and Information Technology

Research output: Chapter in Book/Report/Conference proceeding › Paper in conference proceeding › peer-review

Abstract

A major disadvantage of feedforward neural
networks is still the difficulty to gain insight into their internal functionality. This is much less the case for, e.g., nets that are trained unsupervised, such as Kohonen’s self-organizing feature maps (SOMs). These offer a direct view into the stored knowledge, as their internal knowledge is stored in the same format as the input data that was used for training or is used for evaluation. This paper discusses a mathematical
transformation of a feed-forward network into a SOMlike
structure such that its internal knowledge can be visually
interpreted. This is particularly applicable to networks
trained in the general classification problem domain.

Original language	English
Title of host publication	Proceedings ProRisc?03
Pages	447-452
Publication status	Published - 2003
Event	14th ProRISC Workshop on Circuits, Systems and Signal Processing, 2003 - Veldhoven, The Netherlands, Veldhoven, Netherlands Duration: 2003 Nov 26 → 2003 Nov 27

Conference

Conference	14th ProRISC Workshop on Circuits, Systems and Signal Processing, 2003
Country/Territory	Netherlands
City	Veldhoven
Period	2003/11/26 → 2003/11/27

Subject classification (UKÄ)

Electrical Engineering, Electronic Engineering, Information Engineering

Free keywords

feature maps
selforganizing maps
Neural networks
rule extraction
character recognition.

Cite this

@inproceedings{2d468990d93f4b4cb591ce5fb5056bbb,

title = "Translating feed-forward nets to SOM-like maps",

abstract = "A major disadvantage of feedforward neural networks is still the difficulty to gain insight into their internal functionality. This is much less the case for, e.g., nets that are trained unsupervised, such as Kohonen{\textquoteright}s self-organizing feature maps (SOMs). These offer a direct view into the stored knowledge, as their internal knowledge is stored in the same format as the input data that was used for training or is used for evaluation. This paper discusses a mathematical transformation of a feed-forward network into a SOMlike structure such that its internal knowledge can be visually interpreted. This is particularly applicable to networks trained in the general classification problem domain.",

keywords = "feature maps, selforganizing maps, Neural networks, rule extraction, character recognition.",

author = "vanderZwaag, {B J} and Lambert Spaanenburg and C Slump",

year = "2003",

language = "English",

isbn = "90-73461-39-1",

pages = "447--452",

booktitle = "Proceedings ProRisc?03",

note = "14th ProRISC Workshop on Circuits, Systems and Signal Processing, 2003 ; Conference date: 26-11-2003 Through 27-11-2003",

}

TY - GEN

T1 - Translating feed-forward nets to SOM-like maps

AU - vanderZwaag, B J

AU - Spaanenburg, Lambert

AU - Slump, C

PY - 2003

Y1 - 2003

N2 - A major disadvantage of feedforward neural networks is still the difficulty to gain insight into their internal functionality. This is much less the case for, e.g., nets that are trained unsupervised, such as Kohonen’s self-organizing feature maps (SOMs). These offer a direct view into the stored knowledge, as their internal knowledge is stored in the same format as the input data that was used for training or is used for evaluation. This paper discusses a mathematical transformation of a feed-forward network into a SOMlike structure such that its internal knowledge can be visually interpreted. This is particularly applicable to networks trained in the general classification problem domain.

AB - A major disadvantage of feedforward neural networks is still the difficulty to gain insight into their internal functionality. This is much less the case for, e.g., nets that are trained unsupervised, such as Kohonen’s self-organizing feature maps (SOMs). These offer a direct view into the stored knowledge, as their internal knowledge is stored in the same format as the input data that was used for training or is used for evaluation. This paper discusses a mathematical transformation of a feed-forward network into a SOMlike structure such that its internal knowledge can be visually interpreted. This is particularly applicable to networks trained in the general classification problem domain.

KW - feature maps

KW - selforganizing maps

KW - Neural networks

KW - rule extraction

KW - character recognition.

M3 - Paper in conference proceeding

SN - 90-73461-39-1

SP - 447

EP - 452

BT - Proceedings ProRisc?03

T2 - 14th ProRISC Workshop on Circuits, Systems and Signal Processing, 2003

Y2 - 26 November 2003 through 27 November 2003

ER -

Translating feed-forward nets to SOM-like maps

Abstract

Conference

Subject classification (UKÄ)

Free keywords

Fingerprint

Cite this