This program simulates the Genetic Algorithm's ability to progressively match the given target of 1010101010 from 20 randomly initialized "chromosomes" (represented by integers 0,1 arrays). The algorithm uses a percentage value (Pco 0.9, 0.7, 0.5, 0.3, 0.0) to determine the number of "chromosomes" that will be replicated. The remaining "chromosomes" will undergo a simulated cross-over process, based on the mask 0000011111 (or even split) with random "chromosome" pairings. "Chromosome" fitness values are calculated by the number of gene (array index) matches to the target pattern. Random mutations will occur with every Pco value. Without the random mutation of at least one "gene" in the 20 "chromosome" population, the Pco 0.0 run would never stop, due to 100% of of the "chromosomes" just being replicated. The algorithm will run until at least one chromosome matches the target by displaying the fitness value of 10.
Program will display the initial generated population and all iterations up to the target value, followed by the 2nd and 3rd iterations and then the second to last and last iterations for Pco's 0.7 and 0.0.
Options to Execute Program
From linux command line:
cd genetic_algorithm_cpp
docker build -t ga .
docker run gacd genetic_algorithm_cpp
g++ -o geneticAlgo src/main.cpp
./geneticAlgoIssues Being Worked On
This program lacks quality architecture; however, this is being worked on. The code coupling is very intense and will need to be refactored. The geneticAlgorithm function does take on the qualities of the Template design pattern in how each function is ordered in such a way as to provide accurate data outputs. This program was designed and constructed prior to me having taken Software Engineering or Software Design courses. After refactoring one year later, I see odd arrangements with private and public categorizations; however, the arrangement worked with the sloppy design style. I am also attempting to implement suggestions from "Clean Code" as well. So far, I've given variables better names.
In the event that there are 4 or less iterations before the optimal solution is found, there will be excess generations produced on printout of the "The first generation after initial population"; "The second generation after the initial population"; "The second to last generation"; "The last generation". E.g. if there are only 4 iterations for any run, "The second to last generation" will contain the optimal fitness value of 10 chromosome and "The last generation" will contain an excess generation (which might and might not contain the optimal chromosome since that chromosome only has 10 spots in the chromosome roulette vector. A possible SOLUTION is to include the condition of 'if there are less than 5 generations, then fill the generations after the optimal solution generation with null input'.
Improvements To Be Made
Utilize separate files for better code comprehension. Implement
helper classes for better code organization. Removal of unused functions for
final product. Overall better naming practices will help in understanding what each
variable and function does (e.g., I donw remember what int endSpot = 0; is doing).
Minor Note
Only minor refactors were implemented after having completed this course. The overall state of this program is left for reference of how I developed this program during the class. I do plan on refactoring each function to be smaller (i.e., more legible) and to have better names, which will lead to the removal of the many comments littering the program.