I have 100x73 double mat file as shown in image. I want to remove all zero values from this file and save it to a new mat file. How can I do it? Thank you.

[Chunru]

What do you mean remove all zeros? Will the new data be still a matrix or a vector? Explain what your desired output should be.

[Walter Roberson]

For example do you want it broken down into rectangular blocks, each of which contains no zeros, and put those together into a cell array?

Do you want to convert each item to a cell array entry, but then remove the entries that are {0} so there would be a rectangular cell most of which has scalars but some entries are empty?

Do you want to sparse() the array so that the zeros are not explicitly stored?

[sam]

Hi Walter, I want to keep output as same format as input which is 100x73 double. I only want to remove all zeros and save it to a new mat file.

[the cyclist]

Suppose your matrix were

A = [1 0;
     2 3]

A = 2×2

1 0 2 3

What exactly should the new matrix be? Be specific. Don't just say "A, but without the zeros". A numeric array can't have a "gap" or "hole" in it.

[sam]

Hi some of the row vector has a length of 73, but not all. So I padded it with zero to have same length for all vectors. But I want to unpad it. In that case, some of the row vector will contain blank towards the end of length. Is it not possible?

[Walter Roberson]

Okay, the current input array has a 0 at row 9 column 11. What should be in the output array at row 9 column 11?

Would you happen to be asking for the values at those locations to be interpolated based upon the nearby values ??? Are the 0's "missing values" that you want to interpolate ?

[sam]

Hi Walter, I want to keep those simply blank. That's what I want to do if possibe. So length for each row will change as I don't want to interpolate any values for those.

[Akira Agata]

Is it OK to separate each row with different length?

Like:

A = [1 0 2;...
    2 3 4]

A = 2×3

1 0 2 2 3 4

To

C = {[1 2];...
    [2 3 4]}

C = 2×1 cell array
    {[  1 2]}
    {[2 3 4]}

[sam]

Yes

Answer 1

[Walter Roberson]

It is not possible to have a numeric matrix that contains blanks.

It is not possible to have a numeric matrix that contains a different number of entries in a row or column.

It is possible to write functions that display blanks in place of zeros. For example,

A = rand(5,7); A(randi(35, 1, 4)) = 0

A = 5×7

0.2379 0.2298 0.7207 0.1782 0 0.7270 0.1074 0.7013 0.7682 0.7379 0.6983 0.1428 0.6798 0 0.3510 0 0.6469 0.2545 0.8296 0.5563 0.3290 0.2298 0.5079 0.0852 0.4941 0.9429 0.1410 0.2596 0.2925 0.0965 0.9475 0.6772 0 0.5899 0.3218

C = compose('%6.4f', A)

C = 5×7 cell array
    {'0.2379'}    {'0.2298'}    {'0.7207'}    {'0.1782'}    {'0.0000'}    {'0.7270'}    {'0.1074'}
    {'0.7013'}    {'0.7682'}    {'0.7379'}    {'0.6983'}    {'0.1428'}    {'0.6798'}    {'0.0000'}
    {'0.3510'}    {'0.0000'}    {'0.6469'}    {'0.2545'}    {'0.8296'}    {'0.5563'}    {'0.3290'}
    {'0.2298'}    {'0.5079'}    {'0.0852'}    {'0.4941'}    {'0.9429'}    {'0.1410'}    {'0.2596'}
    {'0.2925'}    {'0.0965'}    {'0.9475'}    {'0.6772'}    {'0.0000'}    {'0.5899'}    {'0.3218'}

C(A == 0) = {'      '};

D = arrayfun(@(R) strjoin(C(R,:), ' '), (1:size(C,1)).', 'uniform', 0);
D = vertcat(D{:});
disp(D)

0.2379 0.2298 0.7207 0.1782        0.7270 0.1074
0.7013 0.7682 0.7379 0.6983 0.1428 0.6798       
0.3510        0.6469 0.2545 0.8296 0.5563 0.3290
0.2298 0.5079 0.0852 0.4941 0.9429 0.1410 0.2596
0.2925 0.0965 0.9475 0.6772        0.5899 0.3218

[Akira Agata]

+1

Another possible solution:

% Sample matrix
A = randi(10, 5, 7); 
A(randi(35, 1, 8)) = 0

A = 5×7

8 0 8 9 3 6 5 10 2 1 6 3 6 10 0 0 0 6 1 10 2 0 3 4 10 6 0 0 10 10 6 6 6 5 1

% Replace 0 with NaN and apply rmmissing for each row
A(A==0) = nan;
C = mat2cell(A, ones(size(A, 1), 1));
C = cellfun(@rmmissing, C, "UniformOutput", false)

C = 5×1 cell array
    {[    8 8 9 3 6 5]}
    {[10 2 1 6 3 6 10]}
    {[       6 1 10 2]}
    {[       3 4 10 6]}
    {[10 10 6 6 6 5 1]}