Exact expected NN error from a dissimilarity matrix (2)
E = NNERROR2(D,M)
Exact computation of the expected 1-nearest neighbor error. D should be a dataset containing a labeled square dissimilarity matrix of a set of objects related to itself. It is not needed that D is symmetric. So NNERROR2(D) ~= NNERROR2(D') is possible.
E = NNERROR2(D)
In this case a set of training set sizes is used to produce a full learning curve. E can be plotted by PLOTE.
There is a similar routine NNERROR1 which is based on the random selection of objects, instead of class-wise.