[scilab-Users] Genetic Algoritm

[michael.baudin at contrib.scilab.org]

 Hi,

 You are right, the "Optimization in Scilab" document does not provide 
 examples of the GA with integer variables. But the demos of Scilab can 
 help - please type :

 demo_gui()

 in the console, and search for the "Optimization and Simulation > 
 Genetic Algorithms > Genetic Algorithms" demo. If you click this, it 
 will execute the demo. The graphics window then contains a button "View 
 Code", which lets you explore the code. The "binary" example is at the 
 end of the script. We used this demo in a training dedicated to 
 optimization we did last year, because it shows the sensitivity of the 
 encoding on the convergence on the algorithm.

 In this example, the real variable is encoded in binary form. This 
 requires to change the objective function, which now gets x as a string 
 which contains a binary coding of the real variable. Before calling the 
 function func (which manages real inputs), we convert the binary string 
 into a real with the convert_to_float function which is provided by the 
 GA toolbox.

 function y=f(x)
   BinLen = get_param(ga_params,'binary_length')
   tmp    = convert_to_float(x, BinLen, Max, Min)
   y      = func(tmp)
 endfunction

 Then there are some particular settings for managing the binary coding 
 of the problem. We first have to set the 'binary_length' option which 
 counts the number of bits used in the encoding and then set the 
 crossover_ga_binary, mutation_ga_binary and coding_ga_binary functions.

 ga_params = add_param(ga_params,'binary_length',8);
 ga_params = set_param(ga_params,'crossover_func',crossover_ga_binary);
 ga_params = set_param(ga_params,'mutation_func',mutation_ga_binary);
 ga_params = set_param(ga_params,'codage_func',coding_ga_binary);

 In another context, working with integer variables may force you to 
 write dedicated crossover_func, mutation_func and codage_func functions. 
 But you may be able to use the current functions with binary encoding : 
 just multiply the output of convert_to_float by 2^8 and you get an 
 integer, uniform in the [0,2^8-1] range.

 -->floor(convert_to_float("11010101",8, 1, 0)*2^8)
  ans  =
     213.
 -->floor(convert_to_float("00000000",8, 1, 0)*2^8)
  ans  =
     0.

 Notice that, in multi-dimensional problems, the string x is the 
 concatenation of the strings in all dimensions. For example, if the 
 problem is in dimension 2, then the 8-bytes string x="11111111" can 
 represent two different 4-bytes values.

 Best regards,

 Michaël

 PS
 The convert_to_float function has no help page :

 http://bugzilla.scilab.org/show_bug.cgi?id=7912

 On Mon, 9 Jul 2012 14:32:46 -0700 (PDT), rcbsimoes 
 <rcbsimoes at hotmail.com> wrote:
> Hi All,
>
> i'm looking for a nice tutorial (with examples) for genetic 
> algorithms.
> specificaly descrete (not continous) input variables.
>
> can same one point me a link?
>
> i've seen the presentation from
> http://www.scilab.org/support/documentation/tutorials  ->  
> Optimization in
> Scilab
>
> this one is nice, but i would like same more tutorials.
>
> greetings from pt, and thanks in advance.
> rui
>
>
> --
> View this message in context:
> http://mailinglists.scilab.org/Genetic-Algoritm-tp4024549.html
> Sent from the Scilab users - Mailing Lists Archives mailing list
> archive at Nabble.com.
>
> --
> To unsubscribe from this mailing-list, please send an empty mail to
> users-unsubscribe at lists.scilab.org
> To check the archives of this mailing list, see
> http://mailinglists.scilab.org/