How to randomize order of functions

I have a function that calls a number of sub-functions. Each of these sub-functions is a test with slight differences but the same basic purpose and a single output. A basic example is given below. What I need to do is have the 4 test functions run in a randomized order each time. I am unsure how to go about it. Any help is much appreciated. Thank you.

function [results] = runTests(testparamters)

    result1 = test1(testparameters);
    result2 = test2(testparameters);
    result3 = test3(testparameters);
    result4 = test4(testparameters);

results = [result1 result2 result3 result4];
end

[Dyuman Joshi]

"What I need to do is have the 4 test functions run in a randomized order each time."

But that wouldn't change the final output of runTests().

Do you want to get the results array with random combination of the result1, result2, result3 and result4 ? (As @Matt has shown below) If this is not what you want, please specify.

[Erik J]

I actually want the results array in order (result 1, result 2, etc), but the functions delivered in random order.

The reason for this is that these tests deliver behavioral experiments to human subjects using different stimuli. Because humans are good at learning, the order that each subject receives the individual tests needs to be random (so it is not the same from subject to subject). This could be done manually, but there are many tests in the protocol and it would be difficult.

Answer 1

[James Tursa]

E.g., Using the cell array approach with randperm:

function [results] = runTests(testparameters)
f = {@test1,@test2,@test3,@test4};
ix = randperm(4);
result1 = f{ix(1)}(testparameters);
result2 = f{ix(2)}(testparameters);
result3 = f{ix(3)}(testparameters);
result4 = f{ix(4)}(testparameters);
results = [result1 result2 result3 result4];
end

If the results are scalars, then this could be shortened to:

function [results] = runTests(testparameters)
f = {@test1,@test2,@test3,@test4};
results = arrayfun(@(x)f{x}(testparameters),randperm(4));
end

And given that this function is now essentially a one-liner, you might just skip the function runTests altogether and simply include that results line in your calling code directly.

That all being said, if there is learning involved from one call to the next (i.e., the test functions are not independent), then you might not want to use the one-liner approach since you cannot assume arrayfun( ) is guaranteed to call the functions in a particular order. In that case, use the explicit order calling code instead (or use an explicit loop).

[Erik J]

Thanks James. This works perfectly.

Answer 2

[Matt]

Hi,

By curiosity in what context do you need to do something like this ?

You can store the functions handles in a cell and call randomly the cells element like this :

x = 1;
N_fun = 3;
[~,random_index] = sort(rand(1,N_fun));
fun_list = {@f1,@f2,@f3};
results = nan(1,N_fun);
for ii=1:N_fun
   results(ii)=fun_list{random_index(ii)}(x); 
end
results

function y =f1(x)
y = 1;
end
function y =f2(x)
y = 2;
end
function y =f3(x)
y = 3;
end

[Erik J]

The reason for this is that these tests deliver behavioral experiments to human subjects using different stimuli. Because humans are good at learning, the order that each subject receives the individual tests needs to be random (so it is not the same from subject to subject). This could be done manually, but there are many tests in the protocol and it would be difficult.

It seems that this approach could work.