Check if matrix contains zero or not and ...

Hi,

I want to check if matrix contains zero or not, and if its zero in a specific column that I pick.. say the first one, then I want to add the last value in that column to the zero ones.

How on earth do I do this in a effective and fast way?

Many thanks...

[Jan]

What have you tried so far and which problems occurred?

Answer 1

[Jürgen]

http://www.mathworks.com/matlabcentral/newsreader/view_thread/255797

[Moshe Flam]

link is broken

Answer 2

[Moshe Flam]

Here's the short answer: Explained further.

    C = M{:,find(sum(M == 0))}; % C are the zero cols indice
    Z = M(:,C) == 0; % Z is an indicator matrix 
                     % marking zeros 
                     % only in columns with zeros
    N = M; % leaving M intact. Creating copy
    N{:,C} = diag(M(end,C)) * Z; 
    % M(end,C) are the last val of the columns with zeros

    % for last non-zero vals, see remark*

_______________________

Detailed Explanation:

To find the zeroes in a column of the matrix M

    colidx = 1; % first column
    col = M(:,colidx);
    idx =  col == 0;

To check if it has zeroes, sum up the result and check it is over zero

    hasZeros = sum(idx(:)) > 0;

So lets work with a matrix, and get the colidxs that have zero:

    zeros =  M == 0;         % indicator: 1 where there's a zero
    sums =   sum(zeros);     % sum per column
    colidxs =   find(sums);  % indice of columns with zeros

    zeroCols = M(:,colidxs); % data of columns with zeros
    zerosOnly = zeros(:,colidxs);  % indicator only for cols with zero

    lastvals = zeroCols(end,:);

We now do a matrix multiplication with the diagonal:

i.e. for the following:

    M = [1,    2,    3,    4;...
         11,  22,    0,   44;...
         111,  0,  333,  444;...
         10,  20,   30,  40];

    colIdxs:  2,3
    lastvals: 20,30

    zeroCols: 0,0 
              0,1  
              1,0
              0,0

    matrix multiply diagonal  
     x       20, 0
             0, 30

    gives: 0,0; 20,0; 0,30;  0,0

Finally, add the result to the data cols using matrix multiplication:

So:

    valsmat = zeroCols * diag(lastvals); % vals matrix
    result = M; % copy of M, leaving M intact
    result{:,colidxs} = M:{:,colidxs} + valsmat; % add the vals matrix

Resulting in:

     M = [ 1,  2,    3,    4;...
          11, 22,   30,   44;...
         111, 20,  333,  444;...
         10,  20,   30,   40];

______

() *Remark:

If you don't want the last value, but rather the last non-zero value of the column, you'll need to create a vector of last values.

See this question on Stack Overflow

So:

    % lastvars = to-be-completed
    N{:,C} = diag(lastvars) * Z;