How to find first '1' in every row

Hi,

I have a matrix,

A = [ 0 0 0 0 0 0 1 1 1 1 0 0; 0 0 0 0 0 1 1 1 1 0 0 0; 0 0 0 0 0 0 1 1 1 1 0 0]

My question is, how to find the first '1' in each row. I want the output will show like this:

B = [7; 6; 7]

Meaning that, for the first row, the number 1 found on column number 7, second row found in column number 6 and so on.

Thank you in advance for any answered

Answer 1

[Matt J]

 [~,B]=max(A,[],2);

[Walter Roberson]

Hah! Good thinking!

[Matt J]

It's interestingly still the only MATLAB feature that I know of which resembles a row-wise find(x,1) operation.

[Kasmayanti Sakaria]

It works. Thank you so much

[ME]

perfect solution! Thanks!

Answer 2

[Jos (10584)]

Not better than using max (for this type of input), but just to show you an alternative:

A = [ 0 0 0 0 0 0 1 1 1 1 0 0; 0 0 0 0 0 1 1 1 1 0 0 0; 0 0 0 0 0 0 1 1 1 1 0 0]
[c, ~] = find(cumsum(A, 2).' == 1)

[ME]

another good solution, but for larger matrices slower

tic; [~, B] = max(A,[],2);               toc;
tic; [c, ~] = find(cumsum(A, 2).' == 1); toc;
size(A)

Elapsed time is 0.014103 seconds.

Elapsed time is 0.153354 seconds.

ans =

1000 12501

[Jos (10584)]

of course it is slower, as it requires three internal loops (cumsum, ==, & find) rather than one :-D

Answer 3

[Walter Roberson]

A = [ 0 0 0 0 0 0 1 1 1 1 0 0; 0 0 0 0 0 1 1 1 1 0 0 0; 0 0 0 0 0 0 1 1 1 1 0 0];

c = sum(cumprod(~A,2),2)+1;

I do sometimes use this techique to do a vectorized "find" if the first or last element along a dimension. However it does force you to think carefully about what value is "reasonable" for the case where there is no match. In this form of the code, a row with no 1 would give you a result which is 1 greater than the number of columns.

Answer 4

[Jan]

Speed test:

A = randi([0,1], 1000, 12501);

tic;      % MAX
[~, B] = max(A,[],2);
toc;
>> 0.007 sec

tic;      % CUMSUM
[c, ~] = find(cumsum(A, 2).' == 1);
toc;
>> 0.160 sec

tic;      % FOR over rows
n = size(A,1);
d = zeros(1, n);
for k = 1:n
    d(k) = find(A(k, :), 1);
end
toc
>> 0.193 sec

tic;     % FOR over columns
At = A.';
n = size(At, 2);
e = zeros(1, n);
for k = 1:n
    e(k) = find(At(:, k), 1);
end
toc
>> 0.119 sec

tic;     % FOR, Accumulate
n = size(A, 2);
d = zeros(size(A,1), 1);
for k = 1:n
    m = A(:, k);
    d(m & ~d) = k;
    if all(d)
        break;
    end
end
toc
>> 0.009 sec  % Worst case: 0.220 sec!!!