In NLTK, the MaxEnt classifier reports multiple accuracy metrics, as shown in the screenshot below. The first set is the accuracy per iteration. The second, printed at the bottom of the screenshot, is reported by calling nltk.classify.accuracy(maxent, test_set), where the maxent variable has been properly initialized. Which of these accuracies are reportable, i.e. in results to show the training model's accuracy? What's the difference between them?
