Based on the theory of complex system, the three kinds of networks are compared firstly in the paper, which are artificial neural network, brain neural network and small-world network. According to the analysis of comparison, it can be observed that the structure of classic artificial neural network is regular, and itpsilas functions have many defects and limit by comparing with brain neural network; the brain neural network is a small-world network; the difference between artificial neural network and small-world network consists in three levels of their complexity: vertices, links and behavior. Then how to reconstruct the classic artificial neural network to small-world neural network is discussed. The small-world neural network could be obtained through the rewiring links with probability p (0<p<1) from artificial neural network. The rewiring probability p should be satisfied with 0<p<0.1 for the best simulation. At last, the paper presents some conclusions.
[1]
Joaquín J. Torres,et al.
Influence of topology on the performance of a neural network
,
2004,
Neurocomputing.
[2]
Mark E. J. Newman,et al.
The Structure and Function of Complex Networks
,
2003,
SIAM Rev..
[3]
Guanrong Chen,et al.
Stability of a neural network model with small-world connections.
,
2003,
Physical review. E, Statistical, nonlinear, and soft matter physics.
[4]
D. Simard,et al.
Fastest learning in small-world neural networks
,
2004,
physics/0402076.
[5]
Duncan J. Watts,et al.
Collective dynamics of ‘small-world’ networks
,
1998,
Nature.
[6]
Guillermo Abramson,et al.
Associative memory on a small-world neural network
,
2003,
nlin/0310033.
[7]
M. Montuori,et al.
Topological approach to neural complexity.
,
2004,
Physical review. E, Statistical, nonlinear, and soft matter physics.