Create an array from a cell array

Given the following cell array

G= {[1  2  1  2  3  4  5  4  5  4  6  3  9  3  9;
     5 10 15 20 25 30 35 40 45 50 55 60 65 70 75],
    [1  1  1  1  2  3  4  4  4  5  4  6  3  6  3  9;
     5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80]}

and consider just G {1,1} (then I will need to iterate to consider also G{1,2})

I want to obtain theese arrays

V1=[1 1 ; 5 0];
V2=[2 2; 10 0];
V3=[3 3 3; 25 0 0];
V4=[4 4 4; 30 0 0]
V5=[5 5, 35 0]
V6=[6; 55]
V7=[9 9, 65 0]

Visualizing G{1,1}

I want to obtain 7 arrays, because 7 is the number of different elements that are inside the first raw of G{1,1}: (1 2 3 4 5 6 9). If the number of diversities is more x, I would like to obtain x arrays.

Then each vector has, in the first raw, the elements repeated as many times as they are inside the first raw of G{1,1}. Instead, in the second raw of each arrays we want to have as many zeros as the columns above, but in the first position (first column, second raw) we want to report the value that is present in the second raw of G{1,1} when the element appear for the first time.

For example, the element 9 appear the first time with value 65, so in V9 we will have 65, as indicated by yellow circles in the figure above.

May someone help me?

Answer 1

[Guillaume]

For just one element of the cell array, eg. G{1}:

[uval, loc1, ids]  = unique(G{1}(1, :));  %get unique values of first row, with location where first occurence is found and unique id for each values
count = accumarray(1, ids)';   %get histogram (count) of each unique value
result = arrayfun(@(v, s, n) [repelem(v, n); s, zeros(1, n-1), uval, G{1}(2, loc1)], count, 'UniformOutput', false);
celldisp(result)

To apply to each element of G just wrap it in a loop:

result = cell(size(G));
for gidx = 1:numel(G)
    [uval, loc1, ids]  = unique(G{gidx}(1, :));
    count = accumarray(ids, 1)';
    result{gidx} = arrayfun(@(v, s, n) [repelem(v, n); s, zeros(1, n-1)], uval, G{idx}(2, loc1), count, 'UniformOutput', false);
end

[luca]

Using the second part, so the iteration, I don't know why but I get this error

Index in position 2 exceeds array bounds (must not exceed 15).

Error in Untitled0 (line 131)
    result{gidx} = arrayfun(@(v, s, n) [repelem(v, n); s, zeros(1, n-1)], uval, G{1}(2, loc1), count, 'UniformOutput',
    false);
    

I think it's because the second cell is 2*16 and no 2*15.

Do you know how to fix this problem?

[Adam Danz]

That's because there's a small typo in the answer.

result{gidx} = arrayfun(@(v, s, n) [repelem(v, n); s, zeros(1, n-1)], uval, G{1}(2, loc1), count, 'UniformOutput', false)
% > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > ^this should be gidx

[Guillaume]

Forgot to replace the {1} by {idx} in the arrayfun, hence the error. Fixed now.

The number of elements in each cell does not matter at all 1,15, 16, 1 million, the code doesn't care.

[Adam Danz]

"g"idx (not idx)

BTW, Guillaume's answer is ~1.6x faster than mine.

[luca]

Hi,

When you have time may you help with this problem ? that is related to your answers.

https://it.mathworks.com/matlabcentral/answers/481921-modify-a-cell-array-referring-to-another-cell-array?s_tid=prof_contriblnk

Thanks in any case

Sincerely, Luca

Answer 2

[Adam Danz]

This solution loops through each element of cell array "G" and produces an output "V" which is a cell array the same size as G. See the notes the follow the code to understand how to retrieve your "V1', "V2", etc variables.

G= {[1 2 1 2 3 4 5 4 5 4 6 3 9 3 9; 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 ],[1 1 1 1 2 3 4 4 4 5 4 6 3 6 3 9; 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 ]}

% Loop through each element of cell array G
V = cell(size(G)); 
for i = 1:numel(G)
    unqVals = unique(G{i}(1,:));          % vector of unique values in 1st row of G{i}
    nReps = sum(unqVals==G{i}(1,:).',1);  % number of repetitions of each unique value (requires r2016b or later)
    row2val = G{i}(2,arrayfun(@(x)find(G{i}(1,:)==x,1,'first'),unqVals)); 
    V{i} = arrayfun(@(x,r,v)[x*ones(1,r);[v,zeros(1,r-1)]],unqVals,nReps,row2val,'UniformOutput',false)
end

"V1" for the first and second elements of G are stored in V{1}{1} and V{2}{1}

"V2" for the first and second elements of G are stored in V{1}{2} and V{2}{2}

"Vn" for the first and second elements of G are stored in V{1}{n} and V{2}{n}

[luca]

Thanks for your help Adam !

Answer 3

[the cyclist]

Another possibility:

[i,j,k] = unique(G{1}(1,:));

numberUniqueG = numel(i);

for ng = 1:numberUniqueG
    ngCount = sum(k==ng);
    V{ng} = zeros(2,ngCount);
    V{ng}(1,:) = ng;
    V{ng}(2,1) = G{1}(2,j(ng));
end

Wrap this in a loop over the elements of G.