SimplePyGA

This project demonstrates a simple Genetic Algorithm (GA) in Python.

System Requirements

This project was built with Python 3.5.2 64-bit and Anaconda 4.1.1 on Windows 10. It does not yet use any Anaconda-specific features and should work with Python 3.5.x on any platform but this has not been tested. It will not work with Python 2.

A Visual Studio 2015 solution and project file are included for those who like to edit using Visual Studio.

Design

Background

GAs simulate the evolution of organisms by natural selection, incorporating the concepts of fitness, cross-breeding and mutation. There is a population of individuals that evolve over many generations until some kind of termination criteria are met. Typically this means that one or more individuals in the population evolves to be some desired level of fitness.

For a given GA application:

• Individuals represent something you want to improve, such as a design or algorithm
• Individuals have a set of characteristics, usually referred to as their chromosomes or genes and usually represented by numeric parameters or lists of values
• Individuals may mutate, in which case one or more genes may randomly change on its own
• Individuals may cross-breed, in which case one or more new individuals result from combining the genes of one or more parents (usually two)
• A population is simply a collection of those individuals which can evolve together by natural selection
• There is some kind of goal or ideal end state which defines the best possible individual
• There is a fitness function that measures how close a given individual is to the goal
• Within a given generation, the fittest Individuals have a better chance of cross-breeding and producing offspring for the next generation, so that over time the average fitness of the population increases and eventually (you hope) you get an individual that matches or gets close to the goal, terminating the GA experiment

Implementation

In SimplePyGA, there are Python classes for:

• Gene (a named attribute with a numeric value and allowed range)
• Individual (a single “creature” with one or more Genes)
• Population (a collection of Individuals)
• Environment (the world that a Population evolves in; settings for each GA run)
• FitnessCalc (a set of methods for determining an Individual’s fitness with respect to a Goal)

Then there is a main script, SimplyPyGA.py, that runs the whole GA.

Credits

The original version of this code, back when it was all in one script file, was based on the article Creating a genetic algorithm for beginners by Lee Jaconson at The Project Spot.

Improvements to the selection algorithm for cross-breeding came from the article Genetic Algorithm for Solving Simple Mathematical Equality Problem by Denny Hermawanto.