Index of /galib-2.4/examples

[ICO]NameLast modifiedSizeDescription

[DIR]Parent Directory  -
[   ]README11-Nov-1999 13:56 12K
[   ]ex1.C11-Nov-1999 13:56 3.4K
[   ]ex2.C13-Dec-1999 23:53 5.1K
[   ]ex3.C11-Nov-1999 13:56 5.0K
[   ]ex4.C11-Nov-1999 13:56 3.7K
[   ]ex5.C11-Nov-1999 13:56 12K
[   ]ex6.C11-Nov-1999 13:56 7.3K
[   ]ex7.C11-Nov-1999 13:56 7.8K
[   ]ex8.C11-Nov-1999 13:56 7.0K
[   ]ex9.C11-Nov-1999 13:56 2.9K
[   ]ex10.C11-Nov-1999 13:56 7.9K
[   ]ex11.C11-Nov-1999 13:56 5.7K
[   ]ex12.C13-Dec-1999 23:55 3.7K
[   ]ex13.C14-Dec-1999 00:12 7.3K
[   ]ex14.C11-Nov-1999 13:56 11K
[   ]ex15.C11-Nov-1999 13:56 4.9K
[   ]ex16.C11-Nov-1999 13:56 7.9K
[   ]ex17.C11-Nov-1999 13:56 3.9K
[   ]ex18.C11-Nov-1999 13:56 5.9K
[   ]ex19.C11-Nov-1999 13:56 7.8K
[   ]ex20.C11-Nov-1999 13:56 5.3K
[   ]ex21.C13-Dec-1999 23:56 7.2K
[   ]ex22.C11-Nov-1999 13:56 9.1K
[   ]ex23.C11-Nov-1999 13:56 3.9K
[   ]ex24.C11-Nov-1999 13:56 7.7K
[   ]ex25.C11-Nov-1999 13:56 1.8K
[   ]ex26.C11-Nov-1999 13:56 12K
[   ]ex27.C11-Nov-1999 13:56 13K
[DIR]gnu/02-Mar-2000 08:03 -
[DIR]graphic/02-Mar-2000 08:03 -
[   ]makefile03-Feb-2000 13:34 2.7K
[   ]makefile.win3214-Dec-1999 23:39 3.0K
[DIR]pvmind/02-Mar-2000 08:03 -
[DIR]pvmpop/02-Mar-2000 08:03 -
[   ]randtest.C11-Nov-1999 13:56 20K
[   ]seed.C11-Nov-1999 13:56 4.2K
[TXT]settings.txt11-Nov-1999 13:56 802
[TXT]smiley.txt11-Nov-1999 13:56 454
[TXT]test_results-linux.txt14-Dec-1999 00:22 27K
[TXT]test_results-winnt.txt14-Dec-1999 13:10 28K
[TXT]tsp_rect_20.txt11-Nov-1999 13:56 132
[TXT]values.txt11-Nov-1999 13:56 35

GAlib: examples Here is what each one of the examples does:

Fill a 2DBinaryStringGenome with alternating 0s and 1s using a SimpleGA.
Generate a sequence of random numbers, then use a Bin2DecChromosome and SimpleGA to try and match the sequence. This example shows how to use the user-data member of genomes in objective functions.
Read a 2D pattern from a data file then try to match the pattern using a 2DBinaryStringGenome and a SimpleGA. This example also shows how to use the GAParametes object for setting genetic algorithm parameters and reading command-line arguments.
Fill a 3DBinaryStringChromosome with alternating 0s and 1s using a SteadyStateGA. This example uses many member functions of the genetic algorithm to control which statistics are recorded and dumped to file.
This example shows how to build a composite genome (a cell?) using a 2DBinaryStringGenome and a Bin2DecGenome. The composite genome uses behaviors that are defined in each of the genomes that it contains. The objective is to match a pattern and sequence of numbers.
Grow a GATreeGenome using a SteadyStateGA. This example illustrates the use of specialized methods to override the default initialization method and to specialize the output from a tree. It also shows how to use templatized genome classes. Finally, it shows the use of the parameters object to set default values then allow these to be modified from the command line. The objective function in this example tries to grow the tree as large as possible.
Identical in function to example 3, this example shows how to use the increment operator (++), completion measure, and other member functions of the GA. It uses a GA with overlapping populations rather than the non-overlapping GA in example 3 and illustrates the use of many of the GA member functions. It also illustrates the use of the parameter list for reading settings from a file, and shows how to stuff a genome with data from an input stream.
Grow a GAListGenome using a GA with overlapping populations. This shows how to randomly initialize a list of integers, how to use the sigma truncation scaling object to handle objective scores that may be positive or negative, and the 'set' member of the genetic algorithm for controlling statistics and other genetic algorithm parameters.
Find the maximum value of a continuous function in two variables. This example uses a GABin2DecGenome and simple GA. It also illustrates how to use the GASigmaTruncationScaling object (rather than the default linear scaling). Sigma truncation is particularly useful for objective functions that return negative values.
Find the maximum value of a continuous, periodic function. This example illustrates the use of sharing to do speciation. It defines a sample distance function (one that does the distance measure based on the genotype, the other based on phenotype). It uses a binary- to-decimal genome to represent the function values.
Generate a sequence of descending numbers using an order-based list. This example illustrates the use of a GAListGenome as an order-based chromosome. It contains a custom initializer and shows how to use this custom initializer in the List genome.
Alphabetize a sequence of characters. Similar to example 11, this example illustrates the use of the GAStringGenome (rather than a list) as an order-based chromosome.
This program runs a GA-within-GA. The outer level GA tries to match the pattern read in from a file. The inner GA tries to match a sequence of randomly generated numbers (the sequence is generated at the beginning of the program's execution). The inner level GA is run only when the outer GA reaches a threshhold objective score.
Another illustration of how to use composite chromosomes. In this example, the composite chromosome contains a user-specifiable number of lists. Each list behaves differently and is not affected by mutations, crossovers, or initializations of the other lists.
The completion function of a GA determines when it is "done". This example uses the convergence to tell when the GA has reached the optimum (the default completion measure is number-of-generations). It uses a binary-to-decimal genome and tries to match a sequence of randomly generated numbers.
Tree chromosomes can contain any kind of object in the nodes. This example shows how to put a point object into the nodes of a tree to represent a 3D plant. The objective function tries to maximize the size of the plant.
Array chromsomes can be used when you need tri-valued alleles. This example uses a 2D array with trinary alleles.
This example compares the performance of three different genetic algorithms. The genome and objective function are those used in example 3, but this example lets you specify which type of GA you want to use to solve the problem. You can use steady state, simple, or incremental just by specifying one of them on the command line. The example saves the generational data to file so that you can then plot the convergence data to see how the performance of each genetic algorithm compares to the others.
The 5 DeJong test problems.
Holland's royal road function. This example computes Holland's 1993 ICGA version of the Royal Road problem. Holland posed this problem as a challenge to test the performance of genetic algorithms and challenged other GA users to match or beat his performance.
This example illustrates various uses of the allele set in array genomes. The allele set may be an enumerated list of items or a bounded range of continuous values, or a bounded set of discrete values. This example shows how each of these may be used in combination with a real number genome.
This example shows how to derive a new genetic algorithm class in order to customize the replacement method. Here we derive a new type of steady-state genetic algorithm in which speciation is done more effectively by not only scaling fitness values but also by controlling the way new individuals are inserted into the population.
The genetic algorithm object can either maximize or minimize your objective function. This example shows how to use the minimize abilities of the genetic algorithm. It uses a real number genome with one element to find the maximum or minimum of a sinusoid.
This example shows how to restricted mating using a custom genetic algorithm and custom selection scheme. The restricted mating in the genetic algorithm tries to pick individuals that are similar (based upon their comparator). The selector chooses only the upper half of the population (so it cannot choose very bad individuals, unlike the roulette wheel selector, for example).
Multiple populations on a single CPU. This example uses the genetic algorithm class called a 'DemeGA'. The genetic algorithm controls the migration behavior for moving individuals between populations. In this example, the island model is used with a stepping-stone migration behavior in which the best individuals from each population migrate to their nearest neighboring population. You can easily modify both the migration algorithm and the population behaviors by deriving a new class from the DemeGA.
Travelling Salesperson Problem. Although genetic algorithms are not the best way to solve the TSP, we include an example of how it can be done. This example uses an order-based list as the genome to figure out the shortest path that connects a bunch of towns such that each town is visited exactly once. It uses the edge recombination crossover operator (you can try it with the partial match crossover as well to see how poorly PMX does on this particular problem).
Deterministic crowding. Although the algorithms built-in to GAlib allow you to do quite a bit of customization, sometimes you'll want to derive your own class so that you can really tweak the way the algorithm works. This example shows one way of implementing the deterministic crowding method by deriving an entirely new genetic algorithm class.
Use this program to verify that the random number generator is generating suitably random numbers on your machine. This is by no means a comprehensive random number testor, but it will give you some idea of how well GAlib's random number generator is working.
graphic ¹
You can learn a great deal by watching the genetic algorithm evolve. This directory contains two examples that show populations of solutions evolving in real time. Both programs use X resources as well as command-line arguments to control their behavior. You can also use a standard GAlib settings file. The programs will compile with either the Motif or athena widget set. Both examples have a simple X windows interface that lets you start, stop, restart, and incrementally evolve a population of indivdiuals. You can see the evolution in action, so it becomes very obvious if your operators are not working correctly or if the algorithm is converging prematurely.

In the first example, the objective function is a continuous function in two variables with concentric rings and a maximal value located in the center. You can choose between 3 different genetic algorithms, 2 different genomes (real or binary-to-decimal), and 4 different functions.

The second example shows solutions to the travelling salesman problem evolving in real time. You can compare three different algorithms: simple, steady-state, and deterministic crowding.
gnu ¹
This directory contains the code for an example that uses the BitString object from the GNU class library. The example illustrates how to incorporate an existing object (in this case the BitString) into a GAlib Genome type. The gnu directory contains the source code needed for the BitString object (taken from the GNU library) plus the two files (bitstr.h and bitstr.C) needed to define the new genome type and the example file that runs the GA (gnuex.C).
pvmind ¹
This directory contains code that illustrates how to use GAlib with PVM in a master-slave configuration wherein the master process is the genetic algorithm with a single population and each slave process is a genome evaluator. The master sends individual genomes to the slave processes to be evaluated then the slaves return the evaluations.
pvmpop ¹
This directory contains code that illustrates a PVM implementation of parallel populations. The master process initiates a cluster of slaves each of which contains a single population. The master process harvests individuals from all of the distributed populations. With a few modifications you can also use this example with the deme GA from example 25 (it uses migration to distribute diversity between pops).
¹ available only in the UNIX distribution
Matthew Wall, 2 January 1996