How to set timeout to a function

Hi everyone, I need to stop the execution of a function if finish a specific time. Specifically the part of the program where i need to implement it is when i call a function to create a regressor:

[trainedModel, validationRMSE] = trainRegressionModelLinear(T, KFolds);

I have left the program running and it never finish the function so i wanted to exit the function and continue the rest of the program. The function works because i change the data "T" that i insert in the function and i get a trainedModel with its value but with long data T i think no works.

There is a example that i need to do but with try- catch doesnt work:

try 
[trainedModel, validationRMSE] = trainRegressionModelLinear(T, KFolds);
catch (*time out finish*)
    validationRMSE = "Error";
    trainedModel = [];
end

I find some information how said that i need to use parallel pool, batch(), parfeval() or a timer but i dont understand what function is useful for my situation. I already use parallel pool with 4 workers because is better to train regressor but to apply the time out no idea.

Thanks so much for the help and sorry if I have explained badly because the english is not my native language.

Answer 1

[Edric Ellis]

You could use parfeval to do this, but I'm not sure how you might fit that in with the rest of your parpool usage. Here's the simplest case where you use wait with a single future:

fut = parfeval(@pause, 0, 20); % Dummy function - you'd use your training function here
ok = wait(fut, 'finished', 4); % Wait for up to 4 seconds for 'fut' to finish
if ~ok
    disp('Did not finish in time.')
else
    disp('Did finish in time.')
end

You could use fetchNext to do something more sophisticated with multiple futures, as that also takes a timeout parameter.

[Javier Checa]

Thank you very much, it helped me.

[Jan Kappen]

I'd like to add, that even without parallel computing toolbox this works:

fut = parfeval(backgroundPool, @pause, 0, 5); % Dummy function - you'd use your training function here

disp('jo ahhah')

ok = wait(fut, 'finished', 4); % Wait for up to 4 seconds for 'fut' to finish
if ~ok
    disp('Did not finish in time.')
else
    disp('Did finish in time.')
end

Note the backgroundPool parameter.

And you can run arbitrary code before calling the wait function, i.e. Matlab remains responsive which is super awesome.

[Vincent]

I get the following error @Jan Kappen:

'parfeval' requires Parallel Computing Toolbox.

[Jan Kappen]

@Vincent then you have an old Matlab version. I do not have Parallel Computing Toolbox installed and it works on R2023a. Be sure to use the backgroundPool.

[Vincent]

@Jan Kappen I am using matlab 2019a

[Jan Kappen]

@Vincent jep, that's too old, see Environment for running code in the background - MATLAB (mathworks.com). R2021b is required.

Answer 2

[Giovanni Soleti]

A similar solution that worked in my case is the following:

% you do what u need

tic

try

s = my_fcn();

catch

disp("out of time");

end

function s = my_fcn()

if toc > timeout

err("you are out of time");

end

% your function ...

% ,,,,,,,,,,,

end