Sensitivity and learning of two digital artificial neural network structures

V. E. DeBrunner; S. C. Li; S. Lewandowsky

Sensitivity and learning of two digital artificial neural network structures

V. E. DeBrunner^*, S. C. Li, S. Lewandowsky

^*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceeding › Conference Contribution (Conference Proceeding)

4 Citations (Scopus)

Abstract

We extend the analysis of parameter sensitivity and interdependence to two digital artificial neural network structures, the backpropagation and ALCOVE. This paper compares the two networks, and we generalize to show that a highly sensitive weight contributes more to the prediction of the network than does an insensitive parameter. This suggests that the information structure of an input pattern can be determined by looking at the sensitivity of the interconnection weights, which has ramifications in network design. Additionally, results from a different set of simulations indicate that information about weight sensitivity and interdependence is predictive of the learning behavior of the networks.

Original language	English
Title of host publication	Proceedings - IEEE International Symposium on Circuits and Systems
Publisher	Institute of Electrical and Electronics Engineers (IEEE)
Pages	445-448
Number of pages	4
Volume	3
Publication status	Published - 1996
Event	Proceedings of the 1996 IEEE International Symposium on Circuits and Systems, ISCAS. Part 1 (of 4) - Atlanta, GA, USA Duration: 12 May 1996 → 15 May 1996

Conference

Conference	Proceedings of the 1996 IEEE International Symposium on Circuits and Systems, ISCAS. Part 1 (of 4)
City	Atlanta, GA, USA
Period	12/05/96 → 15/05/96

Structured keywords

Memory

Cite this

@inproceedings{4c4719a704cf45bc8772e041e86ab649,

title = "Sensitivity and learning of two digital artificial neural network structures",

abstract = "We extend the analysis of parameter sensitivity and interdependence to two digital artificial neural network structures, the backpropagation and ALCOVE. This paper compares the two networks, and we generalize to show that a highly sensitive weight contributes more to the prediction of the network than does an insensitive parameter. This suggests that the information structure of an input pattern can be determined by looking at the sensitivity of the interconnection weights, which has ramifications in network design. Additionally, results from a different set of simulations indicate that information about weight sensitivity and interdependence is predictive of the learning behavior of the networks.",

author = "DeBrunner, {V. E.} and Li, {S. C.} and S. Lewandowsky",

year = "1996",

language = "English",

volume = "3",

pages = "445--448",

booktitle = "Proceedings - IEEE International Symposium on Circuits and Systems",

publisher = "Institute of Electrical and Electronics Engineers (IEEE)",

address = "United States",

note = "Proceedings of the 1996 IEEE International Symposium on Circuits and Systems, ISCAS. Part 1 (of 4) ; Conference date: 12-05-1996 Through 15-05-1996",

}

DeBrunner, VE, Li, SC & Lewandowsky, S 1996, Sensitivity and learning of two digital artificial neural network structures. in Proceedings - IEEE International Symposium on Circuits and Systems. vol. 3, Institute of Electrical and Electronics Engineers (IEEE), pp. 445-448, Proceedings of the 1996 IEEE International Symposium on Circuits and Systems, ISCAS. Part 1 (of 4), Atlanta, GA, USA, 12/05/96.

Sensitivity and learning of two digital artificial neural network structures. / DeBrunner, V. E.; Li, S. C.; Lewandowsky, S.

Proceedings - IEEE International Symposium on Circuits and Systems. Vol. 3 Institute of Electrical and Electronics Engineers (IEEE), 1996. p. 445-448.

Research output: Chapter in Book/Report/Conference proceeding › Conference Contribution (Conference Proceeding)

TY - GEN

T1 - Sensitivity and learning of two digital artificial neural network structures

AU - DeBrunner, V. E.

AU - Li, S. C.

AU - Lewandowsky, S.

PY - 1996

Y1 - 1996

N2 - We extend the analysis of parameter sensitivity and interdependence to two digital artificial neural network structures, the backpropagation and ALCOVE. This paper compares the two networks, and we generalize to show that a highly sensitive weight contributes more to the prediction of the network than does an insensitive parameter. This suggests that the information structure of an input pattern can be determined by looking at the sensitivity of the interconnection weights, which has ramifications in network design. Additionally, results from a different set of simulations indicate that information about weight sensitivity and interdependence is predictive of the learning behavior of the networks.

AB - We extend the analysis of parameter sensitivity and interdependence to two digital artificial neural network structures, the backpropagation and ALCOVE. This paper compares the two networks, and we generalize to show that a highly sensitive weight contributes more to the prediction of the network than does an insensitive parameter. This suggests that the information structure of an input pattern can be determined by looking at the sensitivity of the interconnection weights, which has ramifications in network design. Additionally, results from a different set of simulations indicate that information about weight sensitivity and interdependence is predictive of the learning behavior of the networks.

UR - http://www.scopus.com/inward/record.url?scp=0029721171&partnerID=8YFLogxK

M3 - Conference Contribution (Conference Proceeding)

AN - SCOPUS:0029721171

VL - 3

SP - 445

EP - 448

BT - Proceedings - IEEE International Symposium on Circuits and Systems

PB - Institute of Electrical and Electronics Engineers (IEEE)

T2 - Proceedings of the 1996 IEEE International Symposium on Circuits and Systems, ISCAS. Part 1 (of 4)

Y2 - 12 May 1996 through 15 May 1996

ER -

Sensitivity and learning of two digital artificial neural network structures

Abstract

Conference

Structured keywords

Other files and links

Fingerprint

Cite this