Performance Upgrade of Sequence Detector Evolution Using Grammatical Evolution and Lexicase Parent Selection Method

Quickly designing correct and efficient digital circuits is a crucial need for the electronics industry. Several Electronic Design Automation tools are used for this task. Still, they often lack the diversity of designs that search-based techniques can offer, such as our system producing three different designs for a 5-bit ‘11011’ Sequence Detector. Sequence Detectors are some of the most crucial digital sequential circuits evolved in this work using Grammatical Evolution, a Machine Learning technique based on Evolutionary Computation. Compared to the literature, a reasonably small training data set is used to generate diverse solutions/circuits. A comparison is delivered of the results of the evolved circuits using two different parent selection techniques, tournament selection and lexicase selection. It is shown that the evolved circuits using a small training data set have shown a hundred percent test accuracy on a vast amount of test data sets, and the performance of lexicase selection is much better than tournament selection while evolving these circuits.