Data compression and novelty filtering in retinotopic backpropagation networks

Hanseok Ko; R. H. Baran; Mohammed Arozullah

doi:10.1109/ijcnn.1991.170765

Data compression and novelty filtering in retinotopic backpropagation networks

Hanseok Ko, R. H. Baran, Mohammed Arozullah

School of Electrical Engineering

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

5 Citations (Scopus)

Abstract

Three-layer networks with identical numbers of input and output units were trained using standard backpropagation to reproduce vectors of independent, identically distributed random variables. Data compression was accomplished by virtue of the hidden layer having fewer units than the input or output. The trained nets gave a transparent response to training inputs and a translucent response when anomalies are added to elements of the training set. The reproducing vector closely resembles the unperturbed input. By subtracting the output vector from the input, a novel filter results, since the anomalies are dramatically enhanced in the difference vector.

Original language	English
Title of host publication	91 IEEE Int Jt Conf Neural Networks IJCNN 91
Publisher	Publ by IEEE
Pages	2502-2507
Number of pages	6
ISBN (Print)	0780302273, 9780780302273
DOIs	https://doi.org/10.1109/ijcnn.1991.170765
Publication status	Published - 1991
Event	1991 IEEE International Joint Conference on Neural Networks - IJCNN '91 - Singapore, Singapore Duration: 1991 Nov 18 → 1991 Nov 21

Publication series

Name	91 IEEE Int Jt Conf Neural Networks IJCNN 91

Other

Other	1991 IEEE International Joint Conference on Neural Networks - IJCNN '91
City	Singapore, Singapore
Period	91/11/18 → 91/11/21

ASJC Scopus subject areas

Engineering(all)

Access to Document

10.1109/ijcnn.1991.170765

Cite this

Ko, H, Baran, RH & Arozullah, M 1991, Data compression and novelty filtering in retinotopic backpropagation networks. in 91 IEEE Int Jt Conf Neural Networks IJCNN 91. 91 IEEE Int Jt Conf Neural Networks IJCNN 91, Publ by IEEE, pp. 2502-2507, 1991 IEEE International Joint Conference on Neural Networks - IJCNN '91, Singapore, Singapore, 91/11/18. https://doi.org/10.1109/ijcnn.1991.170765

@inproceedings{354c4a4afc274d86880b346ec9b8d51d,

title = "Data compression and novelty filtering in retinotopic backpropagation networks",

abstract = "Three-layer networks with identical numbers of input and output units were trained using standard backpropagation to reproduce vectors of independent, identically distributed random variables. Data compression was accomplished by virtue of the hidden layer having fewer units than the input or output. The trained nets gave a transparent response to training inputs and a translucent response when anomalies are added to elements of the training set. The reproducing vector closely resembles the unperturbed input. By subtracting the output vector from the input, a novel filter results, since the anomalies are dramatically enhanced in the difference vector.",

author = "Hanseok Ko and Baran, {R. H.} and Mohammed Arozullah",

year = "1991",

doi = "10.1109/ijcnn.1991.170765",

language = "English",

isbn = "0780302273",

series = "91 IEEE Int Jt Conf Neural Networks IJCNN 91",

publisher = "Publ by IEEE",

pages = "2502--2507",

booktitle = "91 IEEE Int Jt Conf Neural Networks IJCNN 91",

note = "1991 IEEE International Joint Conference on Neural Networks - IJCNN '91 ; Conference date: 18-11-1991 Through 21-11-1991",

}

TY - GEN

T1 - Data compression and novelty filtering in retinotopic backpropagation networks

AU - Ko, Hanseok

AU - Baran, R. H.

AU - Arozullah, Mohammed

PY - 1991

Y1 - 1991

N2 - Three-layer networks with identical numbers of input and output units were trained using standard backpropagation to reproduce vectors of independent, identically distributed random variables. Data compression was accomplished by virtue of the hidden layer having fewer units than the input or output. The trained nets gave a transparent response to training inputs and a translucent response when anomalies are added to elements of the training set. The reproducing vector closely resembles the unperturbed input. By subtracting the output vector from the input, a novel filter results, since the anomalies are dramatically enhanced in the difference vector.

AB - Three-layer networks with identical numbers of input and output units were trained using standard backpropagation to reproduce vectors of independent, identically distributed random variables. Data compression was accomplished by virtue of the hidden layer having fewer units than the input or output. The trained nets gave a transparent response to training inputs and a translucent response when anomalies are added to elements of the training set. The reproducing vector closely resembles the unperturbed input. By subtracting the output vector from the input, a novel filter results, since the anomalies are dramatically enhanced in the difference vector.

UR - http://www.scopus.com/inward/record.url?scp=0026287749&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0026287749&partnerID=8YFLogxK

U2 - 10.1109/ijcnn.1991.170765

DO - 10.1109/ijcnn.1991.170765

M3 - Conference contribution

AN - SCOPUS:0026287749

SN - 0780302273

SN - 9780780302273

T3 - 91 IEEE Int Jt Conf Neural Networks IJCNN 91

SP - 2502

EP - 2507

BT - 91 IEEE Int Jt Conf Neural Networks IJCNN 91

PB - Publ by IEEE

T2 - 1991 IEEE International Joint Conference on Neural Networks - IJCNN '91

Y2 - 18 November 1991 through 21 November 1991

ER -

Data compression and novelty filtering in retinotopic backpropagation networks

Abstract

Publication series

Other

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this