Comparing Structured Genetic Algorithms for Artificial Neural Network Generation

A. Molfetas (Australia)


Structured Genetic Algorithm, Neural Networks


Structured Genetic Algorithms have been shown to outperform simple Genetic Algorithms in certain tasks such as the optimization of deceptive and time-variant problems. Though they have been shown to be beneficial, there has not been a conclusive study comparing structured Genetic Algorithms which employ different numbers of genotypic levels. This paper addresses this question. Anal ysis and empirical results resulting from this study suggest that the incorporation of additional layers gives the struc tured Genetic Algorithm a significant performance benefit. This performance benefit diminishes with each additional layer.

Important Links:

Go Back