Link to home
Create AccountLog in
Avatar of wish_C
wish_C

asked on

Positions of Selected Individuals different from thier original position

Am trying to plot the selected individual in This Code and the code for selecting the  top individuals in the population is this:
top_x  = pop(b(1:options.topSize),:)

Open in new window

The code for Generating the initial population is This:
M = 50;   % population size
n = 2;
pop = -50 + 100.*rand(M,n);

Open in new window

. This is done in a for loop. See Here for the entire code

but anytime i look at the plot of the initial population and the selected population, i.e plot for pop and top_x, they look different. The position of the selected ones are different from their position in the initial population. Why is this happening. The concern code is in ID: 39196865.

The plot are outside the loop.

The code for the two plots are here:
figure;
scatter(pop(:,1), pop(:,2), 'x'); and 

figure;
scatter(top_x(:,1),top_x(:,2), 'x');

Open in new window

Avatar of TommySzalapski
TommySzalapski
Flag of United States of America image

You sort eval, but not pop. So it looks like you rearrange the values without rearranging the initial object. This would mean you would be scrambling the data.

Does that seem correct?
Avatar of wish_C
wish_C

ASKER

Yes you are right, TommySzalapski, Now am able to correct the position issue, but the selected individuals are not correct. I need to select points that have best fitness values, but the above code is just selecting points randomly. How can i select points with the best fitness values according to the fitness evaluation.
Avatar of wish_C

ASKER

Best Fitness values are selected base on the eval. That is why i am sorting eval in descending order and select the top 25% as the best individuals. Any help please?
Note that sort can optionally return the order it sorted by so you can use that for the other array.

Try something like this:
[SortIndex, b] = sort(eval, 'descend');
pop = pop(SortIndex)
Avatar of wish_C

ASKER

I have modified This part of the code to include your suggestion:

% Criteria for selecting top 'topSize' individuals based on fitness
% values(eval)
 if strcmp(options.optmType, 'min')
   [void, b] = sort(eval, 'descend');  % maximising
 else
   [void, b] = sort(eval, 'ascend');   % minimising
 end
 pop = pop(void);
 top_x  = pop(b(1:options.topSize),:); % select top individuals
 optm_x = pop(b(1),:);                 % save optimal parameter
 fv     = eval(b(1));                  % save optimal func. value

Open in new window


but it gives the the following error:
Subscript indices must either be real positive integers or logicals.

Error in ==> Imp
 pop = pop(void);

Its seems i am indexing into a matrix or vector using a set of indices that include a number that is not a positive integer or a logical value. when i use

void = round(void)
 pop = pop(void)

Open in new window


It gives me another error Thus:
??? Index exceeds matrix dimensions.

Error in ==> Imp at
 pop = pop(void);

Any idea to solve this?
Sorry. I tried to post a comment, but it must not have gotten through.

What are the values of pop, eval, and void? (set M to something low like 10 please).

You may need to just use a for loop to swap all the values around.
Avatar of wish_C

ASKER

Actually i am using a for loop as you can see in the whole code here Here. I have fix the position issue, the problem is i have calculated the mean and covariance of the top_x (top 25% selected) and sample them again with the same size M as shown here:
 pop = mvnrnd(mu,Sigma,M)

Open in new window


Now i want to plot the result of this sample. Any idea?
ASKER CERTIFIED SOLUTION
Avatar of wish_C
wish_C

Link to home
membership
Create an account to see this answer
Signing up is free. No credit card required.
Create Account
Did my comments not help at all? You can accept multiple comments if something I did helped you figure it out.

Sorry that line didn't work, I must have a different version of MATLAB.
Avatar of wish_C

ASKER

I solve the problem by myself.