The code is divided into two parts and is written in java:

To run and view the code a good software is Dr. Java available for free from:

Once installed one simply has to open both java files above in drjava and click on compile.
In the interactions pane one can then start the program by typing:

            java NetworkEvolution 10000 5 0.2 0.2 10

where the numbers after network evolution are the following parameters in same order as used as arguments in the code:
           n -                     The size of the population to be tested
           hidden -              The number of hidden units each ant has
           mFactor -             The mutation factor, i.e. after each round the offspring will be mutated by a
                                      random number chosen from +/- mFactor
           selectiveStrength -defines what fraction of the population is kept for the generation of the next
                                     offspring, i.e. a selective strength of 0.2 means the top 20% of the generation
                                     get to produce offspring for the next generation
           cycles -                The number of evolutionary cycles the program should run for

The code as it is downloaded will currently print out the cycle number and the mean fitness into the interactions pane.  This can be modified and other info added by adding the line "System.out.println("Your text here " +yourVariableHere);" (without outer apostrophes) in the desired code location.

Web site and all contents © Copyright David Huland 2009, All rights reserved.
Free website templates