0
$\begingroup$

I'm trying to write a genetic algorithm that will take a list of lists of reals in interval [0,1], my "starting population", "score" them according to a fitness function, append the mean and max of these results to a new list, then perform operations to modify the lists and apply the fitness function. This should be for at least 1000 iterations. The code below shows the best attempt I've made so far. First, the input

mystartpop=Table[RandomReal[{0, 1}, 71], 100];

then the algorithm

generator[startingpopulation_ (*e.g., mystartpop*), duration_ (*corresponds to the 
listlength
 of random reals -1, so in this case input is 70*)] := Module[
  {fitnessmeanlist, fitnessmaxlist, i, pop, split, parents, family, 
   mutated, fitscoremax, fitscoremean},
  i = 0;
  fitnessmeanlist = {};
  fitnessmaxlist = {};
  pop[0] = startingpopulation;
  split[i_] := split[i] = selectionGen[pop[i], duration];
  parents[i_] := parents[i] = parentsGen[split[i], 2, 25];
  family[i_] := family[i] = crossover[parents[i]];
  mutated[i_] := mutated[i] = mutateChromosome[family[i], 0.025];
  pop[i_] := pop[i] = mutated[i - 1];
  fitscoremax[i] := fitscoremax[i] = score[#, 70] & /@ pop[i] // Max;
  fitscoremean[i] := 
   fitscoremean[i] = score[#, 70] & /@ pop[i] // Mean;
  While[i < 10,
   fitscoremax[i];
   fitscoremean[i];
   AppendTo[fitnessmaxlist, fitscoremax[i]];
   AppendTo[fitnessmeanlist, fitscoremean[i]];
   split[i];
   parents[i];
   family[i];
   mutated[i];
   i++
   ];
  Return[{fitnessmaxlist, fitnessmeanlist}];
  ]

with the below output

{{-1.29383, fitscoremax$709196[1], fitscoremax$709196[2], 
  fitscoremax$709196[3], fitscoremax$709196[4], fitscoremax$709196[5],
   fitscoremax$709196[6], fitscoremax$709196[7], 
  fitscoremax$709196[8], fitscoremax$709196[9]}, {-20.8473, 
  fitscoremean$709196[1], fitscoremean$709196[2], 
  fitscoremean$709196[3], fitscoremean$709196[4], 
  fitscoremean$709196[5], fitscoremean$709196[6], 
  fitscoremean$709196[7], fitscoremean$709196[8], 
  fitscoremean$709196[9]}}

Why are the fitscoremean and fitscoremax not evaluating for i>0? Is it because split[i], parents[i] etc. aren't working in the While loop? Any solutions are welcomed

EDIT: It seems from further testing that it is split[i] etc. that aren't working, though they work when I run them manually. In words, they should run like this: fitscoremean/max applies a fitness function to 100 lists of 71 reals, determining a fitness value for each, then returning the mean/max value

pop[i] assigns to pop the i-1th output of mutated, or in the 0th iteration it is mystartpop

split[i] returns the top 50% of lists from pop[i], according to their fitness score

parents[i] randomly matches 25 pairs of lists together, with the likelihood of their being selected for pairing determined by their fitness score (repetition is allowed between pairs)

family[i] is a crossover module which converts each list to a single binary string, and randomly chops and pastes it at a crossover point in the string with its paired string, e.g.

{1001001,0101010}

with crossover point at 3 returns

{1001010,0101001}

and finally returns a list including all crossover "children" and all "parents"

mutated[i] randomly changes digits (i.e. "flips" 1s to 0s and vice-versa) in these binary strings, according to a specified probability (e.g. 0.025) then converts lists back to 71 real number elements and returns these lists, which are the values for pop[i+1]

Any help to get this working, and speedily, is appreciated

$\endgroup$
3
  • $\begingroup$ There is far too much code for someone to look through and understand to be able to help you. Suggest you provide a MWE or your question is likely to be closed. $\endgroup$ Apr 20, 2022 at 23:33
  • 1
    $\begingroup$ I think you need fitscoremax[i_] := fitscoremax[i] = score[#, 70] & /@ pop[i] // Max and similar for fitscoremean.; $\endgroup$
    – Carl Woll
    Apr 22, 2022 at 18:45
  • $\begingroup$ @CarlWoll thanks that has indeed fixed it. Only concern is that some of the max values have exceeded the maximum I would have expected (having solved my problem using another method) and it's running exceedingly slow (I need to run 1000 iterations, and 10 iterations took 70 seconds) $\endgroup$
    – J0ta
    Apr 22, 2022 at 19:06

0

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.