Project Status: link
-
Pages
Parallel GA Implementation
The objective of this project is to garner experience working with genetic algorithms (GAs) in a parallel processing environment while developing an understanding of the relationship between computation time and speedup. The UNR Research grid‘s massively-parallel computational power (720 processor cores!) was utilized to facilitate running a simple GA (proportional fitness selection, one-point crossover, and …
Traveling Salesperson Problem (TSP)
A steady-state genetic algorithm described by DeJong was utilized to optimize five traveling salesperson benchmark problems. The normal steady-state GA employs overlap within the populations, cloning subsequent generations of populations, adding new members to the subsequent population, and rejecting poor performers as to keep the population number stable. Selection is accomplished via the roulette-wheel methodology. …
Simple GA tested on DeJong Functions
A simple genetic algorithm with a population of 50 is run for 100 generations with varying parameters against the 5 standard DeJong functions. The algorithm is run with the following parameters: p(xOver) = 0.2, p(mut) = 0.0001 p(xOver) = 0.2, p(mut) = 0.001 p(xOver) = 0.2, p(mut) = 0.01 p(xOver) = 0.67, p(mut) = 0.0001 …