Genetic Algorithm Hello World!
This is a simple project intended to showcase genetic algorithms with a well known example for all new developers; namely the classic "Hello, world!" example!
The application simply "evolves" the string "Hello, world!" from a population of random strings. It is intended to be a gentle introduction into the world of genetic algorithms, using Java, Clojure, Common Lisp, Haskell, Scala, Python and OCaml. The programs themselves are really quite simple, and more complex topics like crossover selection using roulette wheel algorithms, insertion/deletion mutation, etc, have not been included.
I've been working with Genetic Algorithms for a little while now and I stubmled across a C++ implemetation a while ago. I decided to bring it back to life and migrate it to Java with my own enhancements. This is far from ideal code, but it was designed to be a gentle introduction for newcomers to genetic algorithms.
But why the net.auxesia package/namespace?
Auxesia is the greek goddess of spring growth, so when dealing with evolutionary programming like genetic algorithms, the name just seemed to fit. That and I was trying to be witty with my naming, and Dalek just didn't seem right.
The overall architecture for each language is the same. The genetic algorithm is broken up between two logical units: a Chromsome and a Population. In some cases a separate driver is also added, but this is just to keep the logic for the other two components separate and clean.
The Population has 3 key attributes (a crossover ratio, an elitism ratio and a mutation ratio), along with a collection of Chromosome instances, up to a pre-defined population size. There is also an evolve() function that is used to "evolve" the members of the population.
The evolution algorithm is simple in that it uses the various ratios during the evolution process. First, the elitism ratio is used to copy over a certain number of chromosomes unchanges to the new generation. The remaining chromosomes are then either mated with other chromosomes in the population, or copied over directly, depending on the crossover ratio. In either case, each of these chromosomes is subject to random mutation, which is based on the mutation ration metioned earlier.
The crossover algorithm used for mating is a very basic tournament selection algorithm. See Tournament Selection for more details.
Each chromosome has a gene that represents one possible solution to the given problem. In our case, each gene represents a string that strives to match "Hello, world!". Each chromosome also has a fitness attribute that is a measure of how close the gene is to the target of "Hello, world!". This measurement is just a simple sun of the absolute difference of each character in the gene to the corresponding character in the target string above. Each gene is simply a string of 13 ASCII characters from ASCII 32 to ASCII 121 inclusive.
The functions operating on Chromsome include mutate() and mate(), amongst others as necessary for the various language implementations.
The mutate() function will randomly replace one character in the given gene.
The mate() function will take another chromosome instance and return two new chromosome instances. The algorithm is as follows:
- Select a random pivot point for the genes.
- For the first child, select the first n < pivot characters from the first parent, then the remaining pivot <= length characters from the second parent.
- For the second child, repeate the same process, but use the first n < pivot characters from the second parent and the remaining characters from the first parent.
The driver code simply instantiates a new Population instance with a set of values for the population size, crossover ratio, elitism ratio and mutation ratio, as well as a maximum number of generations to create before exiting the simulation, in order to prevent a potential infinite execution.
Depending on the implementation, this code may reside in its own source file.
Take a look at the README files in:
for the specifics for each language.
Each source implementation has unit tests to go along with the source code.
Copyright and License
The MIT License
Copyright © 2011 John Svazic
Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:
The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.