机器学习实战之神经网络用于手写数字识别

有关神经网络（NeuralNetwork）的理论介绍，将在后续章节中进行整理，为了对神经网络有一个直观的理解，本篇博文使用神经网络实现手写数字识别项目，使用 logistic 函数作为激活函数，代码如下，注释比较清楚，不再赘述：

#!/bin/python
#coding=utf-8
# 实现神经网络

import numpy as np

from sklearn.datasets import load_digits
from sklearn.metrics import confusion_matrix, classification_report
from sklearn.preprocessing import LabelBinarizer
from sklearn.cross_validation import train_test_split

#tanh函数
def tanh(x):
    return np.tanh(x)

#tanh函数的导数
def tanh_deriv(x):
    return 1.0 - np.tanh(x)*np.tanh(x)

#逻辑函数
def logistic(x):
    return 1/(1 + np.exp(-x))

#逻辑函数的导数
def logistic_derivative(x):
    return logistic(x)*(1-logistic(x))

#定义神经网络，使用面向对象的特性
class NeuralNetwork:
    #__init__是构造函数，self是指向当前类的指针，相当于java的this指针
    def __init__(self, layers, activation='tanh'):
        """
        :param layers: A list containing the number of units in each layer.
        Should be at least two values
        :param activation: The activation function to be used. Can be
        "logistic" or "tanh"
        """
        if activation == 'logistic':
            self.activation = logistic
            self.activation_deriv = logistic_derivative
        elif activation == 'tanh':
            self.activation = tanh
            self.activation_deriv = tanh_deriv

        self.weights = []
        for i in range(1, len(layers) - 1):
            #layers[i]当前层，layers[i - 1]前一层，layers[i + 1]后一层
            self.weights.append((2*np.random.random((layers[i - 1] + 1, layers[i] + 1))-1)*0.25)
            self.weights.append((2*np.random.random((layers[i] + 1, layers[i + 1]))-1)*0.25)

    #随机抽取一行数据进行一次迭代
    def fit(self, X, y, learning_rate=0.2, epochs=10000):
        X = np.atleast_2d(X)
        #初始化矩阵，值都为1，行数与x一样，列数再加1
        temp = np.ones([X.shape[0], X.shape[1]+1])
        #第一列到倒数第一列都为x
        temp[:, 0:-1] = X  # adding the bias unit to the input layer
        X = temp
        #将list转换成np.array
        y = np.array(y)

        for k in range(epochs):
            i = np.random.randint(X.shape[0])
            #a是随机抽取的一个样本
            a = [X[i]]

            for l in range(len(self.weights)):  #going forward network, for each layer
                #Computer the node value for each layer (O_i) using activation function
                a.append(self.activation(np.dot(a[l], self.weights[l])))  
            error = y[i] - a[-1]  #Computer the error at the top layer
            deltas = [error * self.activation_deriv(a[-1])] #For output layer, Err calculation (delta is updated error)

            #Staring backprobagation
            #从第len(a) - 2层开始，到第0层，每次-1
            for l in range(len(a) - 2, 0, -1): # we need to begin at the second to last layer
                #Compute the updated error (i,e, deltas) for each node going from top layer to input layer

                deltas.append(deltas[-1].dot(self.weights[l].T)*self.activation_deriv(a[l]))
            deltas.reverse()
            for i in range(len(self.weights)):
                layer = np.atleast_2d(a[i])
                delta = np.atleast_2d(deltas[i])
                self.weights[i] += learning_rate * layer.T.dot(delta)

    def predict(self, x):
        x = np.array(x)
        temp = np.ones(x.shape[0]+1)
        temp[0:-1] = x
        a = temp
        for l in range(0, len(self.weights)):
            a = self.activation(np.dot(a, self.weights[l]))
        return a

#handwrite number identify
#下载数据
digits = load_digits()
X = digits.data
y = digits.target
#预处理，将x归一化到0和1之间
X -= X.min()  # normalize the values to bring them into the range 0-1
X /= X.max()

#层数的设定，8*8的图像为64个输入，0-9个数字共10类输出为10，隐藏层有灵活性，一般比输入层多
nn2 = NeuralNetwork([64, 100, 10], 'logistic')
X_train, X_test, y_train, y_test = train_test_split(X, y)
labels_train = LabelBinarizer().fit_transform(y_train)
labels_test = LabelBinarizer().fit_transform(y_test)
#print "start fitting"
nn2.fit(X_train, labels_train, epochs=3000)
predictions = []
for i in range(X_test.shape[0]):
    o = nn2.predict(X_test[i])
    predictions.append(np.argmax(o))
print 'confusion_matrix...'
print confusion_matrix(y_test, predictions)
print 'classification_report...'
print classification_report(y_test, predictions)

输出结果如下：

confusion_matrix...
[[48  0  0  0  0  0  0  0  0  0]
 [ 0 39  0  0  0  1  0  0  2  3]
 [ 0  0 43  0  0  0  0  0  0  0]
 [ 0  0  1 45  0  0  0  1  2  1]
 [ 0  0  0  0 38  0  0  0  2  1]
 [ 0  0  0  0  1 45  1  0  0  2]
 [ 0  1  0  0  0  0 39  0  0  0]
 [ 0  0  0  0  0  0  0 45  3  0]
 [ 0  3  0  0  0  0  0  0 28  2]
 [ 0  0  0  0  2  1  0  0  2 48]]
classification_report...
             precision    recall  f1-score   support

          0       1.00      1.00      1.00        48
          1       0.91      0.87      0.89        45
          2       0.98      1.00      0.99        43
          3       1.00      0.90      0.95        50
          4       0.93      0.93      0.93        41
          5       0.96      0.92      0.94        49
          6       0.97      0.97      0.97        40
          7       0.98      0.94      0.96        48
          8       0.72      0.85      0.78        33
          9       0.84      0.91      0.87        53

avg / total       0.93      0.93      0.93       450