Discussion of article "Genetic Algorithms - It's Easy!" - page 19

New comment
 
Andrey Dik:
You're planning to optimize something, aren't you? So you need to present this "something" in the form of FF, it can be for example the number of profit points on the indicator signals. This is the FF.
I have something to ask, about FF, for example, if it is like the advice given, where does the optimization matter, whether this UGA can save the possibility of profit that has been told or something else, and whether when setting the optimal indicator parameter whether it should be I do virtual testing like the EA example I see, and sorry if this offends, Genetic Algorithms are random, how can we be sure it's optimal, and if possible we enter a good condition to make genetic algorithms optimize parameter indicators, for example, we can take it from historical buy and profit conditions with how many points and genetic algorithms use similar data in the future and run the optimization, is that a good way to use it, I can't think of anything else, if it's wrong please tell me a good use example, if without because and when it is executed then optimal results come out, such as entering random numbers, u Fortunately now, enter the random number again and then lose, and so on, isn't it more of a guesswork.
 
SALEH:
I have something to ask, about FF, for example, if it is like the advice given, where does the optimization matter, whether this UGA can save the possibility of profit that has been told or something else, and whether when setting the optimal indicator parameter whether it should be I do virtual testing like the EA example I see, and sorry if this offends, Genetic Algorithms are random, how can we be sure it's optimal, and if possible we enter a good condition to make genetic algorithms optimize parameter indicators, for example, we can take it from historical buy and profit conditions with how many points and genetic algorithms use similar data in the future and run the optimization, is that a good way to use it, I can't think of anything else, if it's wrong please tell me a good use example, if without because and when it is executed then optimal results come out, such as entering random numbers, u Fortunately now, enter the random number again and then lose, and so on, isn't it more of a guesswork.

Genetic algorithms use random numbers not by themselves, but by using a probability distribution. Since no one knows the future, people always use the probability distribution to make ANY decisions.You can use the full iteration of the parameters, but you won't use them all. You'll need to decide which option to use, won't you? - the result of applying such manual selection in the future is uncertain in advance, that is, it is not deterministic.

 
Andrey Dik:

Genetic algorithms use random numbers not by themselves, but by using a probability distribution. Since no one knows the future, people always use the probability distribution to make ANY decisions.You can use the full iteration of the parameters, but you won't use them all. You'll need to decide which option to use, won't you? - the result of applying such manual selection in the future is uncertain in advance, that is, it is not deterministic.

If so, give me advice on which one is good to optimize, if it is a matter of parameter indicators I slightly disagree because the parameters change frequently. is it better to apply in takeprofit and stoploss or is there another better way. Thank you for your response, and if you have any suggestions, thank you. I'd like to share the results of the experiment and maybe a little help in trying this UGA further.
 
SALEH:
If so, give me advice on which one is good to optimize, if it is a matter of parameter indicators I slightly disagree because the parameters change frequently. is it better to apply in takeprofit and stoploss or is there another better way. Thank you for your response, and if you have any suggestions, thank you. I'd like to share the results of the experiment and maybe a little help in trying this UGA further.

it is difficult to give advice without knowing anything about the trading system

 
Congratulations on the content! However, I was unable to compile the codes or the example mentioned. The error appears to be on the line "ServiceFunction ();"
 
Rafael Ladeia:
Congratulations on the content! However, I was unable to compile the codes or the example mentioned. The error appears to be on the line "ServiceFunction ();"

Many thanks for the feedback in 11 years!))) unfortunately, errors can occur. Looks like it's time for me to update the codes when I only have time for that...

 

I currently have an optimisation problem, and I'm experimenting with this library to see if it can help me. Unfortunately, I'm not quite clear on how to set up the library properly.

For my example, I want to calculate the best (maximum) composition of lot sizes to have the least risk to myself.

To do this, I simply took 6 characters for the test and for these characters the best lot sizes should be found. In the fitness function, I then calculate the expected ROI for those assets and the expected risk at UGA's suggested lot sizes...

If the risk is less than my threshold, I store the expected ROI to maximise in Colony[0][chromos].....

Here is my code:

void CPortfolioRiskMan::GetOptimalWeights()
  {

   double ReplicationPortion_P  = 100.0;
   double NMutationPortion_P    = 10.0;
   double ArtificialMutation_P  = 10.0;
   double GenoMergingPortion_P  = 20.0;
   double CrossingOverPortion_P = 20.0;
   double ReplicationOffset_P   = 0.5;
   double NMutationProbability_P= 5.0;
//-----------------------Variables-------------------------------------
//Preparation of global variables for UGA
   ChromosomeCount=6; //Number of chromosomes in colony
   GeneCount      =6;       //Number of genes
   RangeMinimum   =0.01;    //Minimum of the search range
   RangeMaximum   =1;    //Maximum of the search range
   Precision      =0.01;       //Search step
   OptimizeMethod =2;  //1-minimum, any other-maximum

   ArrayResize(Chromosome,GeneCount+1);
   ArrayInitialize(Chromosome,0);
   Epoch=50;                     //Number of epochs without improvement




//Start of main function of UGA
   UGA
   (
      ReplicationPortion_P, //Portion of replication.
      NMutationPortion_P,   //Portion of natural mutation.
      ArtificialMutation_P, //Portion of artificial mutation.
      GenoMergingPortion_P, //Portion of adoption of genes.
      CrossingOverPortion_P,//Portion of crossing over.
      ReplicationOffset_P,  //Rate of shifting the interval boundaries
      NMutationProbability_P//Probability of mutation of each gene in %
   );



  }



void FitnessFunction(int chromos)
  {
//-----------------------Variables-------------------------------------

   double weight=0.0;
   double ClosePrices[]= {1.21,0.98,1.31,1.22,0.99,0.86};
   string symbols[]= {"EURUSD","GBPUSD","AUDNZD","EURAUD","NZDJPY","GBPJPY"};
   double tickvalues[]= {0.94,0.78,0.56,0.76,0.45,0.94};
   double tickstep=0.00001;
   double Lotsizes[6];
   double av_return=0;

//----------------------------------------------------------------------


   for(int u=1; u<=GeneCount; u++)
     {
      Lotsizes[u-1]=Colony[u][chromos];
      av_return+=Lotsizes[u-1]*tickvalues[u-1]*tickstep*ClosePrices[u-1];
     }

   double risk=CalculateVaR(symbols, Lotsizes);

   if(risk>50)
      Colony[0][chromos]=-10000000.0;
   else
      Colony[0][chromos]=NormalizeDouble(av_return,2);


   AmountStartsFF++;
  }
 
Filip #:

I currently have an optimisation problem, and I'm experimenting with this library to see if it can help me. Unfortunately, I'm not quite clear on how to configure the library properly.

For my example, I want to calculate the best (maximum) composition of lot sizes to have the least risk to myself.

To do this, I simply took 6 characters for the test and for these characters the best lot sizes should be found. In the fitness function, I then calculate the expected ROI for those assets and the expected risk at UGA's suggested lot sizes...

If the risk is less than my threshold, I keep the expected ROI to maximise in Colony[0][chromos].....

Here is my code:

I didn't understand the point of the question, please explain more.

Generally, this is a very old version of the algorithm, it (the version) does not adhere to the more application-friendly algorithm scheme in the current "Population Optimisation Algorithms" articles.

I can recommend using the SDS algorithm or another one from the table, or wait for an updated version of UGA (I don't know if it is appropriate to publish a new look in an article on this old but very powerful algorithm).

I would generally rewrite this article in a new way, with the inclusion of UGA in the ranking table, I don't know - whether it is possible and necessary to do it.

ZY. I think it is not possible to rewrite the existing article, it has already been translated into many languages.

 
wont compile. many errors.
 
gardee005 #:
wont compile. many errors.
Please show me the list of errors. The code is very old, but there is nothing to break even after 14 years.
1...121314151617181920
New comment