Figure 2
From: Computational identification of adaptive mutants using the VERT system
Markov model decision tree. Decoding of the hidden Markov states for each labeled subpopulation occurs as follows. (1) the set of emission symbols O k for a subpopulation is generated from the statistical classifier for all n measurements. (2) The forward Viterbi decoder generates the most likely set of hidden states by choosing the path of maximum likelihood through the system trellis (green lines) based upon the known Markov parameters and O k . (3) The output set X k is assembled from these predictions for all observations.