GALOPPS:

GALOPPS (Genetic Algorithm Optimized for Portability and Parallelism) is a general-purpose parallel GENETIC ALGORITHM system, written in 'C', organized like Goldberg's "Simple Genetic Algorithm". User defines objective function (in template furnished) and any callback functions desired (again, filling in template); can run one or many subpopulations, on one or many PC's, workstations, Mac's, MPP. Runs interactively (GUI or answering questions) or from files, makes file and/or graphical output. Runs easily interrupted and restarted, and a PVM version for Unix networks even moves processes automatically when workstations become busy. (Note: optional GUI requires Tcl/Tk.) 14 example problems included (De Jong Functions, Royal Road, BTSP, etc. )

User may choose:

• problem type (permutation or value-type)

• field sizes (arbitrary, possibly unequal, heeded by CROSSOVER, MUTATION)

• among 7 crossover types and 4 mutation types (or define own)

• among 6 SELECTION types, including "automatic" option based on Boltzmann scaling and Shapiro and Pruegel-Bennett statist. Mechanics stuff

• operator probabilities, FITNESS scaling, amount of output, MIGRATION frequency and patterns,

• stopping criteria (using "standard" convergence STATISTICS, etc.)

• the GGA (Grouping Genetic Algorithm) REPRODUCTION and operators of Falkenauer

GALOPPS allows and supports:

• use of a different representation in each subpopulation, with transformation of migrants

• INVERSION on level of subpopulations, with automatic handling of differing field sizes, migrants

• control over replacement by OFFSPRING, including DeJong crowding or random replacement or SGA-like replacement of PARENTs

• mate selection, using incest reduction

• migrant selection, using incest reduction, and/or DeJong crowding into receiving subpopulation

• optional ELITISM

Generic (Unix) GALOPPS 3.2 (includes 80-pp. manual) is available on ENCORE. For PVM GALOPPS, PC version (different line endings, makefiles), Threaded GALOPPS, and GALOPPS-based 2-level adaptive system, see the MSU GARAGe web site: http://GARAGe.cps.msu.edu/ .

Contact: Erik D. Goodman, <goodman@egr.msu.edu>, MSU GARAGe, Case Center, 112 Engineering Building, MSU, East Lansing, MI 48824 USA.