training.py 文件源码

def trainLimited(self, featureFile, n_datapoints):
    (label_vector, input_vector) = loadData(featureFile)

    trainData, testData, trainLabels, testLabels = \
      cross_validation.train_test_split(input_vector, label_vector, test_size=(0))

    n_totalrows = int((len(label_vector)/n_datapoints))
    for n in range(0, n_totalrows):
      limited_label_vector = trainLabels[0: (n+1) * n_datapoints]
      limited_input_vector = trainData[0: (n+1) * n_datapoints]

      kNNClassifier = neighbors.KNeighborsClassifier(self.n_neighbors, weights='distance')
      kNNClassifier.fit(limited_input_vector, limited_label_vector)

      scores = cross_validation.cross_val_score(kNNClassifier, limited_input_vector, limited_label_vector, cv = 5)
      print '%f on %d datapoints' % ((sum(scores) / len(scores)), len(limited_label_vector))