Too many output argument error while using nested functions?

Hi, I know this is a very common question. I have searched about the issue in multiple places but couldn't find a proper solution to my problem specifically.

So, I have one FEM code that gives multiple outputs, such as:

function [KAPPA, STIFF, BASIS] = RunGMsFEM
........

Then, I have another function that simulates the FEM code (training number) m times,

function [] = TrainData(fun, m, indices)

predefine some variables

...........
    
for k = 1:m
rng(k); % Here I am simulating some of the variables inside fun
[KAPPA, STIFF, BASIS] = fun; % I will run RunGMsFEM function here
    for loop
        .....
    end
end

..........
    
save data as csv

The problem is that when I run the code:

[KAPPA, STIFF, BASIS] = RunGMsFEM

in command window I am not getting any error. But whenever I run the nested function

TrainData(RunGMsFEM, 100, some_indices)

I am getting too many output argument error on the exact same line:

[KAPPA, STIFF, BASIS] = fun

I know MATLAB doesn't let you assign multiple variables at one using vectors (like tuples in Python (a, b) = (3, 5)), and I understand the error but somehow it works perfectly when I run this code in command window. I tried to change my code into one output vector/matrix and then pull out he Z = [KAPPA, STIFF, BASIS] variables from the one output but the problem here all of the outputs have different size. For example: KAPPA is a vector, STIFF and BASIS are matrices with different sizes.

I can run the code in the command window but then I have to predefine all of the variables by hand which is annoying. Is there anyway to use a multiple output function inside another function?

Thank you,

Answer 1

[Matt J]

The code you've posted does not make it clear what problem you are encountering. There doesn't appear to be anything wrong in the code, and you haven't posted any error messages for us to examine.

However, it might help to mention that the Matlab equivalent of Python's (a, b) = (3, 5) would be,

>> [a,b]=deal(3,5)

a =

     3


b =

     5

Or an even more basic way would be,

     a=3; 
     b=5;

[mathLAB]

When I run

[KAPPA, STIFF, BASIS] = RunGMsFEM

in command window I am not getting any error. But whenever I run the nested function

function [] = TrainData(RunGMsFEM, 100, some_indices)

I am getting too many output arguments error on the exact same line:

[KAPPA, STIFF, BASIS] = fun

[Matt J]

It depends on how "fun" is defined, which you haven't shown us. This will work,

fun=@RunGMsFEM;
[KAPPA, STIFF, BASIS] = fun()

[mathLAB]

fun is the RunGMsFEM code, the reason I didn't share it because it is very long and involves over 10 different functions inside. Mainly, it is finite element method and yields a KAPPA vector and STIFF, BASIS matrices.

function [KAPPA, STIFF, BASIS] = RunGMsFEM

[Matt J]

But in the code you have posted, you already have function named RunGMsFEM. So why doesn't your for-loop look like below? Why are you using the alternative name fun at all? And what have you done to tell Matlab that these 2 function names are really the same thing?

for k = 1:m
rng(k); 
[KAPPA, STIFF, BASIS] = RunGMsFEM(); 
    for loop
        .....
    end
end

[mathLAB]

I have posted the generic function definition, sure when I run the function I switch fun = RunGMsFEM, that's why I run the line below:

TrainData(RunGMsFEM, 100, some_indices)

where fun = RunGMsFEM, m =10 and indices = some_indices (predefined variable).

[Matt J]

But as I mentioned a few comments ago, you cannot write fun = RunGMsFEM. You have to write,

fun = @RunGMsFEM

The '@' symbol is important!. Also, you must use parentheses, even if you are passing no arguments to fun,

[KAPPA, STIFF, BASIS] = fun() %The parentheses () are important here!

[mathLAB]

Yes, this works! Thank you. But it doesn't store any of the local variables. I guess I will change the output as the variables I need to store.

Thank you very much!

Answer 2

[Walter Roberson]

[KAPPA, STIFF, BASIS] = fun % I will run RunGMsFEM function here

That is valid syntax; it is not obvious what difficulty you are having?

I tried to change my code into one outpu vector/matrix and then pull out he Z = [KAPPA, STIFF, BASIS] variables from the one output

Have you function emit a cell array ?

Is there anyway to use a multiple output function inside another function?

If you mean along the lines of

fun2(fun1())

where fun1 emits 3 outputs and fun2 expects 3 inputs, then NO, there is no way.

I have not been able to figure out the rules for the number of output arguments, but in the above case fun1() would always be called being told that only one output argument was expected.

[mathLAB]

Sorry, if the question wasn't clear before. So when I run the code:

[KAPPA, STIFF, BASIS] = RunGMsFEM

in command window I am not getting any error. But whenever I run the nested function

TrainData(RunGMsFEM, 100, some_indices)

I am getting too many output arguments error on the exact same line:

[KAPPA, STIFF, BASIS] = fun

[Walter Roberson]

fun does not appear to be the same as runGMsFEM . Perhaps you did something like

fun = RunGMsFEM;
[KAPPA, STIFF, BASIS] = fun

If so, then NO, that is not going to work and there is no way to make it work without changing RunGMsFEM .

If you change RunGMsFEM to emit a cell array (even if only in the case where nargout == 1) then you would be able to do

fun = RunGMsFEM;
[KAPPA, STIFF, BASIS] = fun{:};

[Walter Roberson]

You edited your comment after I replied...

TrainData(RunGMsFEM, 100, some_indices)

If the first position of TrainData is being received as the variable fun, then NO, there is no way to capture multiple outputs from RunGMsFEM there. The closest you can get is to have RunGMsFEM notice nargout == 1 and emit a cell array, and have TrainData pull the parts out of the cell array.

[mathLAB]

Thank you, I will try using a cell array.

Answer 3

[Steven Lord]

The error message is correct. When you run this code:

TrainData(RunGMsFEM, 100, some_indices)

You are calling RunGMsFEM with one output and passing that output into TrainData as the first input argument. Therefore [KAPPA, STIFF, BASIS] = fun is an attempt to assign what's in the array fun into those three variables, and that MATLAB does not allow.

If you want to pass something that will allow TrainData to run RunGMsFEM with the desired number of outputs (instead of calling RunGMsFEM and passing whatver you received from that one call into TrainData) you need to specify the first input as a function handle to RunGMsFEM.

TrainData(@RunGMsFEM, 100, some_indices)

Compare:

x = 1:10;
callFunction = eps
makeFunctionHandle = @eps

"Passing x into" callFunction will error because it will ask for elements 1, 2, ... 10 of callFunction but callFunction only has one element.

callFunction(x)

Passing x into makeFunctionHandle will work and will call the eps function with the vector 1:10 as input.

out1 = makeFunctionHandle(x)
out2 = eps(x)
isequal(out1, out2) % true, calling makeFunctionHandle is like calling eps directly

[mathLAB]

Thank you, it works!