文档介绍：J. King Saud Univ., Vol. 13, Comp. & Info. Sci., pp. 19-47 (. 1421/2001)
CN-Nets for Modeling and Analyzing Neural Networks
Samir M. Koriem
Department of Systems and Computer Engineering,
Faculty of Engineering, Al-Azhar University, Nasr City, Cairo, Egypt
(Received 24 February 1998; accepted for publication 9 February 2000)
Abstract. The concept of colored timed neural Petri nets (CTNPN or Shortly CN-net) which are isomorphic to
neural architectures is proposed. The CN-net technique incorporates the basic features of the neural net and the
modeling capabilities of both colored and timed Petri nets. The essential principles involved in the construction
of the CN-net are discussed in detail. The computation power of the CN-net model is demonstrated through the
“timed reachability graph” (TRG) that is developed from this model. The CN-net is designed to study the
structure properties of the artificial neural networks (ANNs) while the TRG is used for verifying the dynamic
behavior of these networks. Furthermore, the CN-net offers simple and readable model representation making
it easy to design fitting VLSI circuits for complex ANNs. Practical examples are given illustrating the way in
which the CN-net as a novel modeling technique can be employed to simulate the dynamic behavior and
parallel activities of the ANNs.
Keywords: Modeling; Neural networks; Specification; Colored and timed petri nets; Verification.
1. Introduction
The current interest in the development of artificial neural networks (ANNs) is largely
due to their brain-like organizational structure and learning ability. Although still in an
evolu