I find very convenient that Matlab allows for implicit expansion since the 2016 version (for an explanation, see this nice article: https://blogs.mathworks.com/loren/2016/10/24/matlab-arithmetic-expands-in-r2016b/?s_tid=blogs_rc_1).
I was then puzzled to discover that arrayfun on the cpu does not allow for it, while arrayfun when called with gpu arrays does allow for implicit expansion. Below is a MWE to demonstrate this behavior.
Let me quickly explain it: if I have two vectors x with size [2,1] and y with size [1,2], I can calculate the sum x+y and get a matrix 2*2 as intended. This is better than the ugly and more memory-intensive
repmat(x,1,2)+repmat(y,2,1)
Unfortunately this does not work with arrayfun on the cpu!
Since I code both using normal arrays and GPU arrays, I find this different behavior of arrayfun quite misleading. It would be great if Matlab could allow implicit expansion also on arrayfun cpu. When I have large arrays, duplicating dimensions with repmat takes a lot of memory.
%% Demonstration of implicit expansion support in MATLAB and arrayfun
% This script shows that:
% 1) MATLAB supports implicit expansion for standard array operations.
% 2) arrayfun on the GPU supports implicit expansion.
% 3) arrayfun on the CPU does NOT support implicit expansion!!
%
% Implicit expansion allows a 2×1 vector to be added to a 1×2 vector,
% producing a 2×2 matrix.

clear; clc; close all;

% Define test vectors
x = [1; 2]; % Column vector (2×1)
y = [1, 2]; % Row vector (1×2)

%% Implicit expansion using standard MATLAB operations
F1 = myadd(x, y);

%% Implicit expansion using arrayfun on the GPU
F2 = arrayfun(@myadd, gpuArray(x), gpuArray(y));

%% Attempt implicit expansion using arrayfun on the CPU (expected to fail)
try
F3 = arrayfun(@myadd, x, y);
catch ME
fprintf(‘CPU arrayfun error (expected):n%snn’, ME.message);
end

%% Function myadd
function F = myadd(x, y)
% Element-wise addition
F = x + y;
endI find very convenient that Matlab allows for implicit expansion since the 2016 version (for an explanation, see this nice article: https://blogs.mathworks.com/loren/2016/10/24/matlab-arithmetic-expands-in-r2016b/?s_tid=blogs_rc_1).
I was then puzzled to discover that arrayfun on the cpu does not allow for it, while arrayfun when called with gpu arrays does allow for implicit expansion. Below is a MWE to demonstrate this behavior.
Let me quickly explain it: if I have two vectors x with size [2,1] and y with size [1,2], I can calculate the sum x+y and get a matrix 2*2 as intended. This is better than the ugly and more memory-intensive
repmat(x,1,2)+repmat(y,2,1)
Unfortunately this does not work with arrayfun on the cpu!
Since I code both using normal arrays and GPU arrays, I find this different behavior of arrayfun quite misleading. It would be great if Matlab could allow implicit expansion also on arrayfun cpu. When I have large arrays, duplicating dimensions with repmat takes a lot of memory.
%% Demonstration of implicit expansion support in MATLAB and arrayfun
% This script shows that:
% 1) MATLAB supports implicit expansion for standard array operations.
% 2) arrayfun on the GPU supports implicit expansion.
% 3) arrayfun on the CPU does NOT support implicit expansion!!
%
% Implicit expansion allows a 2×1 vector to be added to a 1×2 vector,
% producing a 2×2 matrix.

clear; clc; close all;

% Define test vectors
x = [1; 2]; % Column vector (2×1)
y = [1, 2]; % Row vector (1×2)

%% Implicit expansion using standard MATLAB operations
F1 = myadd(x, y);

%% Implicit expansion using arrayfun on the GPU
F2 = arrayfun(@myadd, gpuArray(x), gpuArray(y));

%% Attempt implicit expansion using arrayfun on the CPU (expected to fail)
try
F3 = arrayfun(@myadd, x, y);
catch ME
fprintf(‘CPU arrayfun error (expected):n%snn’, ME.message);
end

%% Function myadd
function F = myadd(x, y)
% Element-wise addition
F = x + y;
end I find very convenient that Matlab allows for implicit expansion since the 2016 version (for an explanation, see this nice article: https://blogs.mathworks.com/loren/2016/10/24/matlab-arithmetic-expands-in-r2016b/?s_tid=blogs_rc_1).
I was then puzzled to discover that arrayfun on the cpu does not allow for it, while arrayfun when called with gpu arrays does allow for implicit expansion. Below is a MWE to demonstrate this behavior.
Let me quickly explain it: if I have two vectors x with size [2,1] and y with size [1,2], I can calculate the sum x+y and get a matrix 2*2 as intended. This is better than the ugly and more memory-intensive
repmat(x,1,2)+repmat(y,2,1)
Unfortunately this does not work with arrayfun on the cpu!
Since I code both using normal arrays and GPU arrays, I find this different behavior of arrayfun quite misleading. It would be great if Matlab could allow implicit expansion also on arrayfun cpu. When I have large arrays, duplicating dimensions with repmat takes a lot of memory.
%% Demonstration of implicit expansion support in MATLAB and arrayfun
% This script shows that:
% 1) MATLAB supports implicit expansion for standard array operations.
% 2) arrayfun on the GPU supports implicit expansion.
% 3) arrayfun on the CPU does NOT support implicit expansion!!
%
% Implicit expansion allows a 2×1 vector to be added to a 1×2 vector,
% producing a 2×2 matrix.

clear; clc; close all;

% Define test vectors
x = [1; 2]; % Column vector (2×1)
y = [1, 2]; % Row vector (1×2)

%% Implicit expansion using standard MATLAB operations
F1 = myadd(x, y);

%% Implicit expansion using arrayfun on the GPU
F2 = arrayfun(@myadd, gpuArray(x), gpuArray(y));

%% Attempt implicit expansion using arrayfun on the CPU (expected to fail)
try
F3 = arrayfun(@myadd, x, y);
catch ME
fprintf(‘CPU arrayfun error (expected):n%snn’, ME.message);
end

%% Function myadd
function F = myadd(x, y)
% Element-wise addition
F = x + y;
end arrayfun, implicit expansion MATLAB Answers — New Questions

​