Title :
Sensitivity of trained neural networks with threshold functions
Author :
Oh, Sang-Hoon ; Youngjik, Lee
Author_Institution :
Res. Dept., Electron. & Telecommun. Res. Inst., Daejeon, South Korea
fDate :
27 Jun-2 Jul 1994
Abstract :
In this paper, we derive the sensitivity of single hidden-layer networks with threshold functions, called “Madaline”, as a function of the trained weights, the input pattern, and the variance of weight perturbation or the bit error probability of the binary input pattern. The derived results are verified with a simulation of the Madaline recognizing handwritten digits. Our result show that the sensitivity in a trained network is far different from that of networks with random weights
Keywords :
character recognition; error statistics; neural nets; probability; sensitivity; Madaline; bit error probability; handwritten digit recognition; input pattern; sensitivity; single hidden-layer networks; threshold functions; trained neural networks; trained weights; weight perturbation; Cities and towns; Degradation; Error probability; Gaussian noise; Neural network hardware; Neural networks; Neurons; Pattern recognition; Quantization; Telecommunications;
Conference_Titel :
Neural Networks, 1994. IEEE World Congress on Computational Intelligence., 1994 IEEE International Conference on
Conference_Location :
Orlando, FL
Print_ISBN :
0-7803-1901-X
DOI :
10.1109/ICNN.1994.374316
