ga Taking too Much Time wtihout any Results

I am currently working on a project. The main script is using other functions and data from an Excel file as given in the attached .zip file.

The main file is "optTCv2" and my goal is to conduct a ga opitimzation for 20 items. Although, when I run on a single input it gives results, when I use ga it doesn't bring any results or terminate.

Would you please help me identify the reason?

The files are attached.

Answer 1

[Walter Roberson]

demand = table2array (mainData(9:62 , 2:21));

That is a 2D array.

expDemand = expValue(demand, meanDemandResize, stdDemandResize);

The array is passed as the first parameter to expValue()

function [expValue] = expValue(x , mu , sigma)

Where it is received under the name x

for i=1:x

where the 2D array is used as the upper bound for a for loop.

In MATLAB, when you use a non-scalar as a bound in a colon operation, MATLAB uses the first element of the array. The for loop is not iterating over the elements of x in any way: it is equivalent to

for i = 1 : x(1,1)

Then

y1 = cdf('Normal',x-1,mu,sigma);    %% Cumulative Normal Dist. for Integer Value x-1
y2 = cdf('Normal',x,mu,sigma);      %% Cumulative Normal Dist. for Integer Value x
y3 = y2- y1;                        %% Difference between Cum. N. Dist. (x- (x-1))
y4 = y3 .* x;                        %% Expected Value Difference Between * Value

Those steps involving processing all of x but are independent of the loop control variable, and those steps overwrite all of the variables involved completely. So each iteration of the for loop, those lines calculate exactly the same thing.

y5 = y5 + y4;                       %% Calculation of Expected Value

Except there. y5 was initialized to 0 before the for i loop. If you think about what is happening, with y4 being the same each iteration, the end result is going to be the same as if you had not done any for loop and had instead done

if x(1,1) >= 1
    y1 = cdf('Normal',x-1,mu,sigma);    %% Cumulative Normal Dist. for Integer Value x-1
    y2 = cdf('Normal',x,mu,sigma);      %% Cumulative Normal Dist. for Integer Value x
    y3 = y2- y1;                        %% Difference between Cum. N. Dist. (x- (x-1))
    y4 = y3 .* x;                        %% Expected Value Difference Between * Value
    %repeated addition is the same as multiplication
    y5 = y4 .* floor(x(1,1));
else
    y5 = 0;
end

[Walter Roberson]

%for trial
%% Demand of all 20 items
demand = table2array (mainData(9:62 , 2:21));

mainData does not change inside optimizeTotalCostv2 and your computations will run into problems if it changes inside any other function. Therefore it is a waste of time to do that calculation and related computations inside optimizeTotalCostv2 . You should pull out all as many calculations as you can to before that function is called the first time, and you should pass the calculated values into the function using http://www.mathworks.com/help/matlab/math/parameterizing-functions.html

We also recommend that you avoid using global variables, which are the slowest of all kinds of variables, and are rather error prone. Use parameterization instead.

[Fatih]

Thanks for the info. Actually what I am trying to understand is why doesn't ga bring for a long time any results although I have no constraints. The file brought only 3 iterations in 2 hrs time gap.

Regarding your comment, it is very valuable and probably something I would lose a lot of time to figure out. I haven't yet checked the accuracy of my model. But I will revise that part.

Thanks.

[Walter Roberson]

Your expValue() calculations are taking 43475 times longer then they need to, and they are being done every iteration when they do not change and so should have been calculated only once before-hand. Big waste of time there.

[Fatih]

Great thanks. I removed the expVaşlue from the function and put it in the script. So that seems the main problem. I have at least results in acceptable time. Thanks.