Title: Implementation of Generic Systolic Array for Genetic Algorithm
Student: Wang, Liang-Kai 

Abstract

Genetic Algorithms are search mechanisms which employ the principles of
natural selection and mutation to develop solutions to a various search
and optimization problems that are commonly seen in several fields
including artificial intelligence and Computer Aided Design. However,
due to a big set of data, developing a local optimized solution usually
takes a large amount of time.

This project focuses on the implementation of the genetic algorithm back
to the biological evolution. The genetic algorithm from the biological
evolution exhibits a big amount of serialization which cannot be applied
to the systolic array. However, due to the implicit parallelism
characteristics of genetic algorithm, we then can transform the
equations to single assignment equation, localized variable
broadcasting, and systolic array for high-speed parallel computing.

In this project, Java is used to implement the whole algorithm. The goal
is to have a complete user-defined fitness program that can efficiently
and quickly compute the result. However, in this project, fitness
functions are predefined and a final statistic will be reported over
different functions.