Find first and last NaN/Finite element in each row of a large array.

Hello everyone,

I've modified a prexisten array to convert the non-interest values into NaN and now I want to define different regions of the new array throughout the identification of the finite elements.

As te region of interest would be always in the center of the array, each row is composed by x*NaN - y*Finite - z*NaN elements

example_row_1= [NaN NaN 1 2 3 4 NaN NaN NaN]

example_row_2= [NaN NaN NaN 1 2 NaN NaN NaN NaN]

I would like to identify the first and last finite element of each row in order to define the vector of elements comprised between both values.

I've tried something like this:

for i = 1:rows
    for j = 1:columns
    [regions(i,1), regions(i,2)] = find(isfinite(phases(i,:)), 1); % find the first non NaN element in the matrix
    [regions(i,3), regions(i,4)] = find(isfinite(phases(i,:)), 1, 'last'); %find the last non NaN element in the row
    end
end

I would like to extract the (i, j) indexes of each position.

Thank you.

[madhan ravi]

Expected result?

Answer 1

[Ameer Hamza]

Try something like this

example_row_1 = [NaN NaN 1 2 3 4 NaN NaN NaN];
example_row_2 = [NaN NaN NaN 1 2  NaN NaN NaN NaN];
M = [example_row_1; example_row_2];

[r, c] = find(~isnan(M));
idx = accumarray(r, c, [], @(x) {[min(x) max(x)]});
idx = vertcat(idx{:});

Result

>> idx
idx =
     3     6
     4     5

The first column of idx shows the index of the first non-NaN element in that row, 2nd column shows index last non-Value value.

[Gonzalo Molina]

Hello, thank you for the answer. How can I convert the "absolute index" into row,column index to indentify the elemtn in the matrix.

Gonzalo.

[Ameer Hamza]

The idx, in my answer, is not an absolute index. It is the row-wise index. [3 6] is the non-NaN region of the first row. [4 5] is the non-NaN region of the second row. I think you want something like the following

example_row_1 = [NaN NaN 1 2 3 4 NaN NaN NaN];
example_row_2 = [NaN NaN NaN 1 2  NaN NaN NaN NaN];
M = [example_row_1; example_row_2];

[r, c] = find(~isnan(M));
idx = zeros(size(M, 1), 4);
vals = accumarray(r, c, [], @(x) {[min(x) max(x)]});
idx(:,[2 4]) = vertcat(vals{:});
idx(:,[1 3]) = [1:size(M,1); 1:size(M,1)].';

Result

>> idx
idx =
     1     3     1     6
     2     4     2     5

[Gonzalo Molina]

Thank you very much.

Have a nice day.

Answer 2

[madhan ravi]

First = find(~isnan(row1), 1, 'first’)
Last = find(~isnan(row1), 1, 'last')