Hello, I want to use temporary variable value after parfor loop!, how can this achieved??

counter =0;
parfor i = 1:1000
    max=999;
    if max < i
        max = i;
        disp(max);
        assignin('base','counter',max);
    end
     
end
disp(counter);

For example, I want to keep max value but I got zero

[Jan]

This cannot work: 1. you assign the value to the variable counter in the base-workspace, but where does the posted code occur? Inside a function or a script? 2. The order of iterations is undefined in a parfor loop, but the final value depends on the last processed iteration. As consequence your code is equivalent to: randi([1,1000], 1). What do you want to achieve?

[saddam alqasimi]

Thank you Mr, Jan,

Honestly, my code is different. I wrote this example to know the solution that would be given, so that I can apply the same solution with mine.

The code want to apply parfor with is a "script", I have a for loop and I wante to change it to parfor loop. One of the errors I can't deal with was the counter because it's used with other struct variables to assigne them. Also the counter is still needed after the loop.

here is the part of the code, I added the variables initializationlines and the parameters just in case you need to know the type.

%---------- parameters  
radius = 0.01;  
decay = 500;      
fade = 1/decay;
minThreshold = 4
%----------- initialization 
cmcs.Centre = []; 
cmcs.Count = [];  
cmcs.Life = [];   
cmcs.pointsIndices = []; 
outliers = cmcs;
cmcsGraph = graph();
%------------ The loop 
for i = 1:sampleSpeed % 1:100
        if counter<size(dataset,1)
        counter = counter+1;
        dataPoint = dataset(counter,1:end-1);
        [cmcs,cmcsGraph,outliers] = ...
            Assign(cmcs,outliers,dataPoint,radius,minThreshold,cmcsGraph,counter);
        end
     end

Is it possiable, or what Im trying to do is not right?

Answer 1

[Walter Roberson]

Create cmsc as a struct array of full size. Assign to cmsc(i) as appropriate. Also keep a logical vector used(i) =true where the condition is met. After the parfor cmsc = cmsc(used)

That graph might be more of a problem. Your code structure tends to hint that you are using information from previous iterations in updating the graph.

[saddam alqasimi]

Thanks a lot Mr Walter Roberson,

So, is it better to parallel the outer loop to avoid the problem of updating the graph? or using graph is the problem?

[Walter Roberson]

You deleted the update you had made showing two loops, so I don't have a context for this.

[saddam alqasimi]

Sorry, I thought I dublicated my comment

while counter<size(dataset,1)
    tic;
    time = time +1;
     for i = 1:sampleSpeed 

        if counter<size(dataset,1)
        counter = counter+1;
        dataPoint = dataset(counter,1:end);
        [cmcs,cmcsGraph,outliers] = ...
            Assign(cmcs,outliers,dataPoint,radius,minThreshold,cmcsGraph,counter);
        end

     end
    callingTimesAMean(time) = mean(callingTimesA(counter-sampleSpeed+1:counter));
    timeTableAMean(time) = mean(timeTableA(counter-sampleSpeed+1:counter));
    [cmcs,outliers,cmcsGraph] = Kill(cmcs,outliers,fade,cmcsGraph);
    
    if mod(time,datalogSpeed)==0
        dataLogged(time/datalogSpeed).graph = cmcsGraph;
        dataLogged(time/datalogSpeed).cmcs = cmcs;
        dataLogged(time/datalogSpeed).outliers = outliers;
        
        [targets,outputs] = ExtractPointOutputsTargets(dataLogged,time/datalogSpeed,dataset);
        dataLogged(time/datalogSpeed).targets = targets;
        dataLogged(time/datalogSpeed).outputs = outputs;
    end
    t=toc;
    ts=[ts t];
end

[Walter Roberson]

You are calling Assign() and Kill() to update cmcs in ways that are not obvious. As outside observers, we would have to assume that the results are dependent upon the order of iterations.

[saddam alqasimi]

Yeah, It depends on the iterations but I think it doesn't have to be ordered by iterations as much as it should go through all data-points in the dataset.