Methods perplexity. Lower perplexity on the testing data indicates a better generalization capability. Here we held out 20% of collection for the testing purpose and used the remaining 80% to train the model, in accordance with 5-fold cross-validation.
Wang et al. BMC Bioinformatics 2009 10(Suppl 1):S55 doi:10.1186/1471-2105-10-S1-S55