How can I detect NaN values in a matrix or vector?

How do I identify NaN values?

[Fawad Chandio]

Hello I doing facial recognition attendance system for my final year project I already completed some task but now the problem is that how to recognitions faces...? Help me sir

Answer 1

[Doug Hull]

By definition, NaN is not equal to any number, not even NaN itself. Therefore there are two ways to detect NaN values:

% Generate sample data
 x = rand(1, 10);
 x(x > 0.5) = NaN;
% Find NaN values in two different ways
 y1 = isnan(x) ; 
 y2 = (x ~= x) ;

For speed purposes the use of isnan() tends to be 20%-30% faster. Here's a test snippet if you want to see the comparison:

A = rand(1000); %Random 1000x1000 matrix
A(rand(size(A))>.75) = nan; %Populate with NaNs

t1 = 0;  %time of isnan
t2 = 0;  %time of ~=

for ii = 1:100

      tic
      idx1 = isnan(A); 
      t1 = t1+toc;
      tic
      idx2 = A~=A;
      t2 = t2 + toc;
end

ratio = t2/t1;  %ratio of ~= to isnan
isequal(idx1,idx2)  %Insure same results

%{
ratio = 1.2179
ans = 1
%}

[From the MATLAB FAQ of Ancient Times...]

Answer 2

[Bozydar Wrona]

%num is a vector/matrix, if searching for elements different than NaN then:

num(isnan(num)==0)

%otherwise:

num(isnan(num)==1)