Evaluate several objective functions in parallel (parfor)

Hello Matlab community,

I am looking for help because I have an issue with parfor and optimization algorithms.

I have an optimization written as follows:

[x,fval] = patternsearch(@ObjFunc,start_par,[],[],[],[],min_par,max_par,@constraints,opts);

I need to optimize the parameters to separately fit an ensemble of curves, stored in an array. To speed the calculation up, I would like to use "parfor".

In my ObjFunc, I have the following code:

load('file.mat');
exp_curve = file(:,li);

where "li" is the loop variable.

Working with a normal "for cycle", I would pass "li" as global. Global variables do not work while using "parfor", and I haven't yet found a way to pass the loop variable to my ObjFunc.

Do you have suggestion that can help me?

Best regards.

Answer 1

[Walter Roberson]

You should avoid loading data inside an objective function, as that function will be called many times. You should instead http://www.mathworks.com/help/matlab/math/parameterizing-functions.html parameterize

That could include,

filestruct = load('file.mat', 'file');
file = filestruct.file;
for li = 1 : whatever
    func{i} = @(x) Objfunc(x, file(:,li));
end
parfor li = 1 : whatever
    patternsearch(func{i}, ....)
end

or you could

filestruct = load('file.mat', 'file');
file = filestruct.file;
parfor li = 1 : whatever
    patternsearch( @(x) Objfunc(x, file(:,li)), ...)
end

[Maria Angela Agizza]

Hello, thanks a lot for the answer, it works fine!

Best regards.