#TravSales

Forked from BradHeintz's parallelised genetic algorithm attack on the Traveling Salesman problem. See Brad's related blog post.

The project has been converted into a hierarchical genetic algorithm with migration. The implementation starts from a single job, chaining a configurable number of hierarchical jobs.

The algorithm has also been generalised - the classes in the generalalgorithm package can be implemented with a problem other than TSP. The implementation for TSP can be found in the travsales package.

##Requirements

• Java 7
• Maven
• Hadoop

##To compile

From the project directory run 'mvn clean compile assembly:single'. This will create a single jar file in the target directory.

##To run

Start your Hadoop cluster, then run './hadoop -jar projectdirectory/target/TravSales...jar'.

As the algorithm runs output for each generation will be sent to standard out and written to the file travsales_populations/result. The final best chromosome will be written to the file travsales/finalresult.

##Command line parameters

• -populationSize or -ps : set the population size (must be a multiple of 10).
• -maxNumGenerations or -g : set the maximum number of generations the algorithm iterates over.
• -numHierarchyLevels or -h : set the number of hierarchy levels.
• -filepath or -f : set the filepath for the results of each generation to be stored in as well as for the temporary files created by each hierarchy level.
• -mutationChance or -mc : set the chance of mutation in the range [0.0, 1.0]. Must be a float.
• -migrationRate or -mr : change the rate of migration, i.e. epoch length.
• -migrationPercentage or -mp : change the percentage of individuals that migrate. Must be a float in the range [0.0, 1.0].
• -survivorProportion or -sp : chance the proportion of individuals that survive each iteration. Must be a float in the range [0.0, 1.0].

##License

This software is released under the Apache License, version 2.0.