Too many input arguments error when cell2mat and mean functions are used

I have five cluster data saved in my cell array iniclu. Each row of cell array contains varying number of columns such as 23 27 26 17 7 so that the shortage of columns are marked as [] to the max of 27. Now i want to process cell array row one by one to find mean or centroid of each cluster. My usage is shown in code, but it throws error.

count = [23 27 26 17 7]
iniclu = cell(1,1); %later i built the cluster data as shown in fig 
for i = 1:5
    bb = [];
    bb = cell2mat(iniclu{i,1:count(i)});
    centroids(i,:) = mean(bb,1);
end

But it throws too many input arguments error, hence i could not get expected op as centroids having 5x2 means data. I have attached content of iniclu data. Pls clarify!

[Chuguang Pan]

The cell2mat function only need one cell array input argument. I think the code should be:

bb=cell2mat(iniclu(i,:));

[NALLARASU KRISH]

Thank you. But your answer partially solves the problem when i try to calculate mean. Can't i get cell2mat output as ,for ex first cell row, 23x2 size instead of 1x46 double?

[Walter Roberson]

no, you have different sizes of cells; you can only put them together as a vector (unless you are willing to pad the entries)

Answer 1

[Stephen23]

V = [23;27;26;17;7];
N = numel(V);
C = nan(N,2);
for k = 1:N
    M = vertcat(iniclu{k,1:V(k)}); % comma-separated list
    C(k,:) = mean(M,1);
end

https://www.mathworks.com/matlabcentral/discussions/tips/847976-tutorial-comma-separated-lists-and-how-to-use-them

[NALLARASU KRISH]

Wow ..its a magic!

Answer 2

[VBBV]

bb = cell2mat(iniclu(i,1:count(i)));

Use a ( ) in place of { }

[VBBV]

If you use { } the result is not a cell, so you don't need cell2mat function to convert

[NALLARASU KRISH]

Thank you. But I could not successfully calculate mean. i think my iniclu data structure of x y coordinates for the five clusters is not consistent with mean function of matlab.

[VBBV]

@NALLARASU KRISH, you can get the cell2mat output as 23 x 2 if you reshape

bb = cell2mat(iniclu(i,1:count(i)));
bb = reshape(bb,[],2);
centroids{i} = mean(bb,1);  % The 5th row unequal data size compared to remaining rows

[NALLARASU KRISH]

Your idea reshaped ...but with a trick in it... ha ha ha! Anyway i learnt abt reshape, thank you. Reshape took each 2 values from row and placed it in same column, ie reshape fills up column-wise. Hence pair of x coor and y coor appeared in same column instead of at col 1 and col 2, thus leading to false mean value or centroid.

[VBBV]

@NALLARASU KRISH, ok, try the below

bb = cell2mat(iniclu(i,1:count(i)));
bb = reshape(bb,2,[]).';
centroids{i} = mean(bb,1);  % The 5th row unequal data size compared to remaining rows

[NALLARASU KRISH]

Still not yielded. let me show what happens for each style of statement. Assume 5 pairs

60 52

69 44

85 19

5 99

56 68

it is in iniclu as [60 52] [69 44] [85 19] [5 99] [56 68]

bb = cell2mat(iniclu(1,1:5));

o/p

60 52 69 44 85 19 5 99 56 68

bb = reshape(bb,[],2);

o/p

60 19

52 5

69 99

44 56

85 68

bb = reshape(bb,[],2).';

o/p

60 52 69 44 85

19 5 99 56 68

It looks like small problem, but demands more time!

[VBBV]

@NALLARASU KRISH you have used the below line incorrectly, it should be as shown below

bb = reshape(bb,2,[]).';

[NALLARASU KRISH]

OMG! This is working, you are really great. Thank you!! I accept this answer too. Sorry for my carelessness. Could you explain the difference bw bb = reshape(bb,2,[]).'; and bb = reshape(bb,[],2).'; that did the magic, @VBBV ?

[Steven Lord]

Look at the sizes of the outputs and the arrangement of data in those outputs.

bb = 1:10

bb = 1×10

1 2 3 4 5 6 7 8 9 10

bb1 = reshape(bb, 2, []).'

bb1 = 5×2

1 2 3 4 5 6 7 8 9 10

bb2 = reshape(bb, [], 2).'

bb2 = 2×5

1 2 3 4 5 6 7 8 9 10