MATLAB: How to best determine if something is “NaN”

comparisonnan

I am calculating the ILU factorization of a matrix (repeatedly over a large number of iterations) and at some point norm(U\(L\A)-eye(n),'fro') evaluates to "NaN" (and therefore GMRES does not work). I would like to catch the code immediately before it calls GMRES when this occurs. What is the best way to ask Matlab to stop when norm(U\(L\A)-eye(n),'fro') is NaN?

Best Answer

See documentation for isnan

Related Solutions

MATLAB: Eigenvalues and orthonormal eigenvectors

Yes there is. Look at the eig function.

MATLAB: How to find position of something

You can use the 'find' and 'diff' functions as well as 'isnan' to accomplish that. Let A be your column vector of elements with possible NaNs among them.

 m = 5; % The minimum length of consecutive NaNs to record
 f = find(diff([false;isnan(A);false]));
 B = [f(1:2:end-1),f(2:2:end)-1]; % Get all start and end indices for NaN consecutive sequences
 B = B(B(:,2)-B(:,1)+1>=m,:); % Eliminate all with fewer than m NaNs

The first column of array B contains the start indices and the second column the end indices for consecutive sequences of at least five NaNs.

Best Answer

Related Solutions

MATLAB: Eigenvalues and orthonormal eigenvectors

MATLAB: How to find position of something

Related Question