Previously, researchers have applied neural networks to model and optimize microwave systems that are difficult to analyze and design. The performance of these neural network approaches, as reported is encouraging. In general, the design and training of an appropriate neural network for the problem on hand is hard and time consuming. However, once an appropriate neural network is created, its use is fast and efficient. Genetic algorithms, which mimic the principle of natural selection to find a solution to an optimization problem on the other hand, are easy to apply. The drawback of existing genetic algorithms, however, is their long execution time. In this paper we describe our development of a neural network which is designed and trained by a genetic algorithm. The benefit of this approach is that we can highly simplify the preparation phase of the neural network and still enjoy its easy and fast production capability. The details of our approach as it is adapted to microwave systems is explained, Examples of applying this approach to the impedance matching of microwave circuits is presented.
[1]
Michel Nakhla,et al.
A neural network modeling approach to circuit optimization and statistical design
,
1995
.
[2]
John Holland,et al.
Adaptation in Natural and Artificial Sys-tems: An Introductory Analysis with Applications to Biology
,
1975
.
[3]
Gary S. May,et al.
Advantages of plasma etch modeling using neural networks over statistical techniques
,
1993
.
[4]
P. L. Werner,et al.
Extraction of equivalent circuits for microstrip components and discontinuities using the genetic algorithm
,
1998
.
[5]
M. Vai,et al.
Microwave circuit analysis and design by a massively distributed computing network
,
1995
.
[6]
M. Vai,et al.
Automatic impedance matching with a neural network
,
1993,
IEEE Microwave and Guided Wave Letters.
[7]
Rob A. Rutenbar,et al.
Simulated annealing algorithms: an overview
,
1989,
IEEE Circuits and Devices Magazine.