Odtnhj

I have a very interesting problem, I've started developing a genetic learning algorithm and have succeeded on doing so. its a simple GA designed to find a phrase by randomly selecting characters to store into strings and using the standard selection and mutation methods to progress until the it has the final answer, and sometimes this works perfectly.

However, sometimes there's one character incorrect.
I think this is due to the sorting algorithm being slow.
this is what i have so far

This is the loop code

while (!word.Get_found())

 generation++;
 word.Calculate_fitness();
 word.Selection(); //selection
 word.Crossover(); //crossover

 system("cls");
 std::cout << "Generation: " << generation << " Highest fitness: " << word.get_fittest() << " with string: " << word.get_item() << "n";

This is the code for the fitness function

void Guess_word::Calculate_fitness()// calculates fittness based on guess 

word against matching string;

 for (int i = 0; i < population.size(); i++)
 
 population.at(i).second = 0;
 for (int j = 0; j < population.at(i).first.size(); j++)
 
 if (population.at(i).first.at(j) == Phrase.at(j))
 
 population.at(i).second += 1;//calculate fitness
 
 
 if (population.at(i).second == Phrase.size() && population.at(i).first == Phrase)
 
 found = true;
 

 

And this is the selection function

void Guess_word::Selection()//determine highest fitness of population and make them parents

 //i hate stable sort....
 //it indicates to sort in pairs and keep them together
 std::sort(population.begin(), population.end(), (auto &a, auto &b) return a.second > b.second; );


 //select two random parent from mating pool
 parents.clear();

 parents.push_back(population.at(0));
 parents.push_back(population.at(1));

The population entities are in vector pairs with strings and ints representing the guess and fitness respectively.
after debugging the code i found that the population does indeed contain the correct guess but with the wrong fitness, i think the sorting algorithm moves the ints faster than the paired strings. meaning that during the fitness function it selects an item as the answer that is one character incorrect however with the correct fitness moved from another vector entity.

I've tried using stable sort and moving the algorithm around to see if timing is a problem. however, no dice.
is there a way to either make the program wait for the sort to complete (which is inefficient in terms of time) or a way to either make the sort faster or implement a faster custom sorting algorithm which would be much more efficient especially on older hardware.

Any suggestions would be greatly appreciated!

The issue was simply, the code doing a cross over and storing it back in position 0 of the population making it change randomly just before the final result was displayed