Returns unique state assignment from a (row-wise) weight matrix
Converts a probabilistic cluster assignment to a unique cluster assignment using the
'argmax' rule:state of row i is assigned as the position of the maximum in that row (ties are broken at random).
'sample' rulestate of row i is sampled from the discrete distribution where probabilities equal the weight vector in row i
weight_matrix2states(weight.matrix, rule = c("argmax", "sample"))weight.matrix |
an N \times K matrix |
rule |
how do we choose the state given the weight
matrix. |
WW <- matrix(runif(12), ncol = 3) WW <- normalize(WW) WW weight_matrix2states(WW) weight_matrix2states(WW, "sample") # another 'sample' is in general different from previous conversion unless WW is # a 0/1 matrix weight_matrix2states(WW, "sample")
Please choose more modern alternatives, such as Google Chrome or Mozilla Firefox.