Mean of a cell array with different cell sizes?

Mean of a cell array with different cell sizes?

% This works:
a = {[1 3 4 5];[7 7 8 2];[5 4 1 9]}

a = 3×1 cell array
    {[1 3 4 5]}
    {[7 7 8 2]}
    {[5 4 1 9]}

mean(cell2mat(a),2)

ans = 3×1

3.2500 6.0000 4.7500

% but this does not work:
a = {[1 3 4 5];[7 7 8 2];[5 4 1]}

a = 3×1 cell array
    {[1 3 4 5]}
    {[7 7 8 2]}
    {[  5 4 1]}

mean(cell2mat(a),2)

Error using cat
Dimensions of arrays being concatenated are not consistent.

Error in cell2mat (line 83)
m{n} = cat(1,c{:,n});

Answer 1

[Dyuman Joshi]

If the data in the cell array is compatible for concatenation, concatenate them and use mean for the specific dimension -

a = {[1 3 4 5];[7 7 8 2];[5 4 1 9]}

a = 3×1 cell array
    {[1 3 4 5]}
    {[7 7 8 2]}
    {[5 4 1 9]}

b = cat(1,a{:})

b = 3×4

1 3 4 5 7 7 8 2 5 4 1 9

m = mean(b, 2)

m = 3×1

3.2500 6.0000 4.7500

If the data in the cell array is not compatible for concatenation, the best approach would be to pre-allocate the output and use a for loop. You could use cellfun() but that is just a for loop in disguise.

[Sim]

Gosh, you were very fast to reply and meanwhile I modified my question a bit.. :-)

Btw, Thanks a lot, very kind!

...but what if the cells sizes are different, like in the following example?

a = {[1 3 4 5];[7 7 8 2];[5 4 1]}

a = 3×1 cell array
    {[1 3 4 5]}
    {[7 7 8 2]}
    {[  5 4 1]}

mean(cell2mat(a),2)

Error using cat
Dimensions of arrays being concatenated are not consistent.

Error in cell2mat (line 83)
m{n} = cat(1,c{:,n});

[Dyuman Joshi]

As I mentioned in my answer - pre-allocate the output and use a for loop.

a = {[1 3 4 5];[7 7 8 2];[5 4 1]}

a = 3×1 cell array
    {[1 3 4 5]}
    {[7 7 8 2]}
    {[  5 4 1]}

n = numel(a);
c = zeros(n,1);
for k=1:n
    c(k) = mean(a{k});
end
c

c = 3×1

3.2500 6.0000 3.3333

[Sim]

Got it, many thanks!

(also with cellfun as you said)

a = {[1 3 4 5];[7 7 8 2];[5 4 1]}

a = 3×1 cell array
    {[1 3 4 5]}
    {[7 7 8 2]}
    {[  5 4 1]}

cellfun(@mean,a,'uni',false)

ans = 3×1 cell array
    {[3.2500]}
    {[     6]}
    {[3.3333]}

[Dyuman Joshi]

Speed comparison -

%Generate random data
vec = randi([25 50], 1, 50000);
s = sum(vec)

s = 1873054

arr = randi([1e3 1e4], 1, s);

%Data in the form of cell array
data = mat2cell(arr, 1, vec);


f1 = @(x) funfor(x);
f2 = @(x) funcell(x);


F1 = @() f1(data);
F2 = @() f2(data);

%First check whether output is equal or not
isequal(F1(), F2())

ans = logical
   1

%Timings
fprintf('Time taken by the for loop = %f seconds', timeit(F1))

Time taken by the for loop = 0.025663 seconds

fprintf('Time taken by cellfun = %f seconds', timeit(F2))

Time taken by cellfun = 0.103503 seconds

function c = funfor(x)
n = numel(x);
c = zeros(1,n);
for k=1:n
    c(k) = mean(x{k});
end
end


function c = funcell(x)
c = cellfun(@mean, x);
end

[Dyuman Joshi]

@Sim, Using the name-argument pair will give the output as a cell array.

a = {[1 3 4 5];[7 7 8 2];[5 4 1]}

a = 3×1 cell array
    {[1 3 4 5]}
    {[7 7 8 2]}
    {[  5 4 1]}

cellfun(@mean,a)

ans = 3×1

3.2500 6.0000 3.3333

[Sim]

ah ok, even better! Super thank you!