|
|
Evolutionary Circuit Design by Means of Genetic Programming
Synák, Maroš ; Hurta, Martin (oponent) ; Bidlo, Michal (vedoucí práce)
This thesis investigates the application of genetic programming (GP) for designing circuits, focusing on sinusoidal and square oscillators, and tone discriminators, using Python and PySpice. It aims to recreate aspects of John Koza's foundational work in this field. The primary goal is to assess if GP can generate advanced electronic designs more effectively than traditional methods, adapting Koza's genetic operations—selection, crossover, mutation—to modern circuit design challenges. The methodology involves developing GP models to simulate evolutionary circuit design, evaluated through controlled experiments. These experiments test the models' ability to evolve from basic to complex configurations that meet specific electronic functions. This study not only revisits but also modifies Koza’s methods, incorporating strategies more reliant on an initial embryo setup to guide the evolution process in generating targeted designs. Additionally, the study investigates recent methodologies utilized in similar applications to enhance the adaptability and efficiency of GP. The results demonstrate that while GP can effectively support and enhance electronic circuit design for sinusoid oscillators and tone discriminators, its application to square oscillator generation encounters limitations and severe problems. This highlights areas for potential improvement in genetic diversity and algorithm refinement.
|
host ::
RSS.