def testSupervisedTraining(self): """Test parameter fitting Come up with a good non-trivial way to test your training function You can use the given icecream dataset or make up your own Add your implementation """ seq1 = Instance(label = ['odd', 'even', 'odd', 'even', 'odd'], data = [3, 2, 1, 4, 1]) seq2 = Instance(label = ['even', 'even', 'odd', 'odd', 'even'], data = [2, 4, 1, 3, 2]) seq3 = Instance(label = ['even', 'even', 'odd', 'odd', 'odd'], data = [1, 2, 3, 4, 3]) seq4 = Instance(label = ['odd', 'odd', 'even', 'even', 'even'], data = [4, 3, 4, 1, 2]) instances = [seq1, seq2, seq3, seq4] hmm = HMM() hmm.train(instances) mystery = Instance(data = [2, 1, 3, 4, 2, 2, 1, 3]) labels = hmm.classify_instance(mystery) self.assertEqual(labels, ['even', 'odd', 'odd', 'even', 'even', 'even', 'odd', 'odd'], 'NOOO')
print compute_PER(references, hypotheses) hmm.print_parameters() """ #unsupervised training #for mode in ['uniform', 'random', 'sneaky']: for mode in ['sneaky']: hmm = HMM() hmm._be_prepared_for_baum_welch(training_set, 'sneaky') hmm.baum_welch_train(training_set) #Experiment x.2 hypotheses2 = [hmm.classify_instance(x) for x in training_set] references2 = [x.old_label for x in training_set] print "Over Training Set:" print compute_PER(references2, hypotheses2) #Experiment x.3 hypotheses = [hmm.classify_instance(x) for x in test_set] references = [x.label for x in test_set] print "Over Test Set:" print compute_PER(references, hypotheses) hmm.print_parameters()
