Complexity Analysis of Neural Network Using First Category Orthogonal Weight Functions

Abstract:

To describe the performances of a new kind of neural network, the complexity for training neural network using orthogonal weight functions is analysed. The full adders are used as the neurons of the neural networks, and the weight functions are orthogonal functions. We derive the relationships of the iteration time with the number of input dimensions, output dimensions and training patterns. Finally, some simulation examples verified the theoretical results obtained in this paper.

• Periodical: Advanced Materials Research (Volumes 989-994)
• Pages: 4437-4440
• DOI: https://doi.org/10.4028/www.scientific.net/AMR.989-994.4437
• Online since: July 2014
• Authors: Dai Yuan Zhang*, Jian Hui Zhan
• Keywords: Algorithm Complexity, Neural Network (NN), Orthogonal Approximation, Weight Function

References

[1] Martin T. Hagan，Howard B. Demuth. Neural Network Design [M]. Beijing: Machinery Industry Press, (2006).

[2] Zhang Daiyuan. New Theories and Methods on Neural Networks[M]. Beijing: Tsinghua University Press, (2006).

[3] Li Qingyang, Wang Nengchao, Yi Dayi. Numerical Analysis (4th edition) [M]. Wuhan: Huazhong University of Science and Technology Press, (2006).

[4] Shiow-Shung Yang, Ching-Shiow Tseng. An Orthogonal Neural Network for Function Approximation [J]. Systems, Man, and Cybernetics, Part B: Cybernetics, IEEE Transactions on, 1996, 26(5): 779-785.

DOI: 10.1109/3477.537319

[5] Thomas H. Cormen, Charles E. Leiserson. Introduction to Algorithms [M]. Cambridge: The MIT Press, (2009).