神经网络MPLClassifier分类

 1 # -*- coding: utf-8 -*-
 2 """
 3 Created on Fri Aug 24 14:38:56 2018
 4 
 5 @author: zhen
 6 """
 7 import gzip
 8 import pickle
 9 import numpy as np
10 from sklearn.neural_network import MLPClassifier
11  # 加载数据
12  # 设置编码，解决异常：UnicodeDecodeError: ‘ascii‘ codec can‘t decode byte 0x90 in position 614: ordinal not in range(128)
13 with gzip.open("E:/mnist.pkl.gz") as fp:
14     training_data, valid_data, test_data = pickle.load(fp, encoding=‘bytes‘)
15 x_training_data, y_training_data = training_data
16 x_valid_data, y_valid_data = valid_data
17 x_test_data, y_test_data = test_data
18 classes = np.unique(y_test_data)
19 # 将验证集和训练集合并
20 x_training_data_final = np.vstack((x_training_data, x_valid_data))
21 y_training_data_final = np.append(y_training_data, y_valid_data)
22 # 设置神经网络模型参数
23 # 使用solver=‘lbfgs‘，拟牛顿法，需要较多的跌点次数
24 lbfgs = MLPClassifier(solver=‘lbfgs‘, activation=‘relu‘, alpha=1e-4, hidden_layer_sizes=(50, 50), random_state=1, max_iter=10, verbose=10, learning_rate_init=0.1)
25 # 使用solver=‘adam‘，基于随机梯度下降的优化算法，准确率较低
26 adam = MLPClassifier(solver=‘adam‘, activation=‘relu‘, alpha=1e-4, hidden_layer_sizes=(50, 50), random_state=1, max_iter=10, verbose=10, learning_rate_init=0.1)
27 # 使用solver=‘sgd‘，基于梯度下降的自适应优化算法，分批训练数据，效率高，准确性高，建议使用
28 sgd = MLPClassifier(solver=‘sgd‘, activation=‘relu‘, alpha=1e-4, hidden_layer_sizes=(50, 50), random_state=1, max_iter=10, verbose=10, learning_rate_init=0.1)
29 # 使用不同算法训练模型
30 lbfgs.fit(x_training_data_final, y_training_data_final)
31 adam.fit(x_training_data_final, y_training_data_final)
32 sgd.fit(x_training_data_final, y_training_data_final)
33 # 预测
34 lbfgs_predict = lbfgs.predict(x_test_data)
35 adam_predict = adam.predict(x_test_data)
36 sgd_predict = sgd.predict(x_test_data)
37 print(lbfgs_predict)
38 print("*******************************************")
39 print(adam_predict)
40 print("@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@")
41 print(sgd_predict)
42 print("!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!")
43 # 评估模型
44 print(lbfgs.score(x_test_data, y_test_data))
45 print("===========================================")
46 print(adam.score(x_test_data, y_test_data))
47 print("-------------------------------------------")
48 print(sgd.score(x_test_data, y_test_data))
49 
50 # 输出正确结果
51 print(y_test_data)