The original number of classes: the matrix will define a combining from K to (K-1) classes.
l1
Label of one of the two classes to be combined.
l2
Label of the other class to be combined.
Value
If z is a vector (length K) whose kth entry is the probability that an observation belongs to the kth class in a K-classes classification, then combiM %*% z is the vector (length K-1) whose kth entry is the probability that the observation belongs to the kth class in the K-1-classes classification obtained by merging classes l1 and l2 in the initial classification.
References
J.-P. Baudry, A. E. Raftery, G. Celeux, K. Lo and R. Gottardo (2010). Combining mixture components for clustering. Journal of Computational and Graphical Statistics, 19(2):332-353.