MPNN.py 文件源码

def find_nearest_instance_subprocess(test_instance_start_index, test_instance_end_index,\
                                      classified_results):
    # print test_instance_start_index, test_instance_end_index
    for test_instance_index in range(test_instance_start_index, test_instance_end_index):
        # find the nearest training instance with cosine similarity
        maximal_cosine_similarity = -1.0
        maximal_cosine_similarity_index = 0
        for training_instance, training_instance_index in\
         zip(training_data_instances, range(len(training_data_instances))):
            # compute the cosine similarity
            # first, compute the inner product
            inner_product = np.inner(test_data_instances[test_instance_index], training_instance)
            # second, normalize the inner product
            normalized_inner_product = inner_product / test_data_lengths[test_instance_index]\
             / training_data_lengths[training_instance_index]
            if normalized_inner_product > maximal_cosine_similarity:
                maximal_cosine_similarity = normalized_inner_product
                maximal_cosine_similarity_index = training_instance_index
        classified_results[test_instance_index] =\
         training_data_labels[int(maximal_cosine_similarity_index)]