Find all values in an array neighbored on both sides by NaN.

I have an array that looks like this:

A = [1 2 4 2 NaN 2 4 NaN 6 NaN NaN 9 5 NaN];

I would like to find all values in A that have a NaN immediately adjacent on both sides. In the example above, the only value that meets this criteria is the 6. I can do it in a loop like this:

ind = false(size(A)); 
for n = 2:length(A)-1
    if isnan(A(n-1)) && isnan(A(n+1))
        ind(n)=true; 
    end
end

Is there a more elegant way to do this?

Answer 1

[Sean de Wolski]

A = [1 2 4 2 NaN 2 4 NaN 6 NaN NaN 9 5 NaN]; 
A(conv(double(isnan(A)),[1 0 1],'same')==2)

[Sean de Wolski]

Note, the for-loop is plenty elegant. It could be made more elegant by calling isnan once and then indexing into it on each iteration.

idx = isnan(A)
n = numel(A);
for ii = 1:n-2
 if idx(ii) && idx(ii+2)
   A(ii+1)
 end
end

[Chad Greene]

Convolution, of course! Thanks Sean.