implicit expansion for arrayfun

I want to do something similar to bsxfun, but for more than 2 arrays. For example

arrayfun(@f, A,B,C,D)

where A is a m-by-n matrix, B is a scalar, C is a m-by-1 vector, D is a 1-by-n vector. Right now arrayfun would just throw an error.

I wonder if there is something that could first check whether they are compatible, and then apply repmat so that they all become m-by-n. I couldn't find anything, but hopefully someone is better at this.

Edit:

It appears that arrayfun will do this when input is a gpuArray

"B = arrayfun(FUN,A1,...,An) applies FUN to the elements of the arrays A1,...,An, so that B(i,j,...) = FUN(A1(i,j,...),...,An(i,j,...)).The function FUN must take n input arguments and return a scalar. The nonsingleton dimensions of the inputs A1,...,An must all match, or the inputs must be scalar. Any singleton dimensions or scalar inputs are virtually replicated before being input to the function FUN."

https://www.mathworks.com/help/parallel-computing/gpuarray.arrayfun.html

I will dig more into this.

[Rik]

I suspect you will have to write something yourself. What did you try?

Answer 1

[Yi-xiao Liu]

Apparently arrayfun is a build-in function and I cannot access its source code. Here is what I come up with.

[a,b,c,d]=ImplicitExpansion(nan,[1,2,3],[4;5;6],inf)

a = 3×3

NaN NaN NaN NaN NaN NaN NaN NaN NaN

b = 3×3

1 2 3 1 2 3 1 2 3

c = 3×3

4 4 4 5 5 5 6 6 6

d = 3×3

Inf Inf Inf Inf Inf Inf Inf Inf Inf

function varargout=ImplicitExpansion(Array)

arguments (Repeating)
    Array {mustBeNumericOrLogical}
end


dim=cellfun(@ndims,Array);
dim(cellfun(@isvector,Array))=1;
dim(cellfun(@isscalar,Array))=0;

[~,seq]=sort(dim,'descend');


for i=2:numel(seq)
    Array{seq(i)}=Array{seq(i)}+zeros(size(Array{seq(i-1)}));
end
Array{seq(1)}=Array{seq(1)}+zeros(size(Array{seq(end)}));


varargout=Array;
end

It will throw 'MATLAB:dimagree' error when size was incompatible. Also only work with 2D numerical arrays.

Answer 2

[Matt J]

A=rand(10,5); B=3; C=rand(10,1); D=rand(1,5);

args={A,B,C,D};

sz=cellfun(@size,args,'uni',0);
maxsz=max(vertcat(sz{:}));

newargs= cellfun(@(c)repmat(c, maxsz./size(c)), args,'uni',0);
[A,B,C,D]=deal(newargs{:});

whos A B C D

  Name       Size            Bytes  Class     Attributes

  A         10x5               400  double              
  B         10x5               400  double              
  C         10x5               400  double              
  D         10x5               400  double