Figure 2
From: Improved functional prediction of proteins by learning kernel combinations in multilabel settings
![Figure 2](http://media.springernature.com/full/springer-static/image/art%3A10.1186%2F1471-2105-8-S2-S12/MediaObjects/12859_2007_Article_1896_Fig2_HTML.jpg)
Graphical representation of hidden Markov models. Graphical representation of hidden Markov models. Left: standard 4-class discriminant analysis with 4 Gaussian emission probabilities. denotes the hidden random variable "functional class", represents the observed data. The nodes (red circles) represent the K values of (in this example K = 4). Right: Additional layer with random variable , "subcellular localization class", and two conditionally independent observed variables 1, 2. The different widths of the arrows symbolize different transition probabilities.