《Web安全之机器学习入门》笔记：第十四章 14.3 神经网络 识别验证码

本小节是通过示例如何通过神经网络识别验证码（mnist数据集）。

1、数据集

MNIST实际上是计算机视觉入门数据集，如下图14-6所示：

代码如下所示，将60000个样本作为训练集，剩下的作为测试样本，为了将数据标准化，将其data除以255，保证范围在0-1之间

mnist = fetch_mldata("MNIST original")
# rescale the data, use the traditional train/test split
X, y = mnist.data / 255., mnist.target
X_train, X_test = X[:60000], X[60000:]
y_train, y_test = y[:60000], y[60000:]

因为数据集未下载成功，可以通过如下地址下载，并放到指定目录中

https://raw.githubusercontent.com/amplab/datascience-sp14/master/lab7/mldata/mnist-original.mat

指定目录确定的方法为

from sklearn.datasets.base import get_data_home
print(get_data_home())

程序在运行过程中，报错如下

C:\ProgramData\Anaconda3\lib\site-packages\sklearn\utils\deprecation.py:85: DeprecationWarning: Function fetch_mldata is deprecated; fetch_mldata was deprecated in version 0.20 and will be removed in version 0.22. Please use fetch_openml.
  warnings.warn(msg, category=DeprecationWarning)

 这是因为原书配套github使用的函数到现在已经不使用了，fetch_mldata在新版的sklearn里已经被弃用了，新版的是openml可以通过如下方法修改：，

from sklearn.datasets import fetch_openml
mnist = fetch_openml('mnist_784')

2、神经网络训练

将神经元设置为50个，层数设置为1层，使用sgd法

mlp = MLPClassifier(hidden_layer_sizes=(50,), max_iter=10, alpha=1e-4,
                    solver='sgd', verbose=10, tol=1e-4, random_state=1,
                    learning_rate_init=.1)

mlp.fit(X_train, y_train)

3.完整代码

print(__doc__)

import matplotlib.pyplot as plt
from sklearn.datasets import fetch_mldata
from sklearn.neural_network import MLPClassifier
from sklearn.datasets.base import get_data_home
print (get_data_home())
mnist = fetch_mldata("MNIST original")
# rescale the data, use the traditional train/test split
X, y = mnist.data / 255., mnist.target
X_train, X_test = X[:60000], X[60000:]
y_train, y_test = y[:60000], y[60000:]

mlp = MLPClassifier(hidden_layer_sizes=(50,), max_iter=10, alpha=1e-4,
                    solver='sgd', verbose=10, tol=1e-4, random_state=1,
                    learning_rate_init=.1)

mlp.fit(X_train, y_train)
print("Training set score: %f" % mlp.score(X_train, y_train))
print("Test set score: %f" % mlp.score(X_test, y_test))

fig, axes = plt.subplots(4, 4)
# use global min / max to ensure all weights are shown on the same scale
vmin, vmax = mlp.coefs_[0].min(), mlp.coefs_[0].max()
for coef, ax in zip(mlp.coefs_[0].T, axes.ravel()):
    ax.matshow(coef.reshape(28, 28), cmap=plt.cm.gray, vmin=.5 * vmin,
               vmax=.5 * vmax)
    ax.set_xticks(())
    ax.set_yticks(())

plt.show()

4、运行结果

如下所示，准确率达到了97%左右，效果还是不错的

Iteration 1, loss = 0.32212731
Iteration 2, loss = 0.15738787
Iteration 3, loss = 0.11647274
Iteration 4, loss = 0.09631113
Iteration 5, loss = 0.08074513
Iteration 6, loss = 0.07163224
Iteration 7, loss = 0.06351392
Iteration 8, loss = 0.05694146
Iteration 9, loss = 0.05213487
Iteration 10, loss = 0.04708320


Training set score: 0.985733
Test set score: 0.971000

可视化效果

 在运行过程中报警如下：

C:\ProgramData\Anaconda3\lib\site-packages\sklearn\neural_network\multilayer_perceptron.py:566: ConvergenceWarning: Stochastic Optimizer: Maximum iterations (10) reached and the optimization hasn't converged yet.
  % self.max_iter, ConvergenceWarning)

这是因为迭代次数不够是还是没达到最佳拟合，只要在配置过程中增加迭代次数即可。