搭个简单的神经网络

用Python纯撸一个识别手写数字的简单神经网络

import numpy as np

def sigmoid(x):
    # s形 激活函数
    s = 1 / (1 + np.exp(-x))
    return s

class NeuralNetword:
    """神经网络"""
    def __init__(self, inputnodes, hiddennodes, outputnodes, lr):
        # 三层 输入层 隐藏层 输出层
        self.inodes = inputnodes
        self.hnodes = hiddennodes
        self.onodes = outputnodes
        self.lr = lr

        # 权重
        self.wih = np.random.normal(0.0, pow(self.hnodes, -0.5), (self.hnodes, self.inodes))
        self.who = np.random.normal(0.0, pow(self.onodes, -0.5), (self.onodes, self.hnodes))

        # 激活函数
        self.ac_fun = lambda x: sigmoid(x)

    def train(self, inputs_list, targets_list):
        """训练"""
        inputs = np.array(inputs_list, ndmin=2).T
        targets = np.array(targets_list, ndmin=2).T

        hidden_inputs = np.dot(self.wih, inputs)
        hidden_outputs = self.ac_fun(hidden_inputs)

        final_inputs = np.dot(self.who, hidden_outputs)
        final_outputs = self.ac_fun(final_inputs)

        output_errors = targets - final_outputs

        hidden_errors = np.dot(self.who.T, output_errors)

        self.who += self.lr * np.dot((output_errors * final_outputs * (1.0 - final_outputs)),
                                     np.transpose(hidden_outputs))

        self.wih += self.lr * np.dot(hidden_errors * hidden_outputs * (1.0 - hidden_outputs),
                                     np.transpose(inputs))

    def query(self, inputs_list):
        """查询"""
        inputs = np.array(inputs_list, ndmin=2).T
        hidden_inputs = np.dot(self.wih, inputs)
        hidden_outputs = self.ac_fun(hidden_inputs)

        final_inputs = np.dot(self.who, hidden_outputs)
        final_outputs = self.ac_fun(final_inputs)

        return final_outputs

trains = None
cnn = NeuralNetword(784, 100, 10, 0.1)
with open('mnist_train.csv', 'r') as f:
    trains = f.readlines()
    for t in trains:
        values = t.split(",")
        inputs = (np.asfarray(values[1:]) / 255.0 * 0.99) + 0.01
        targets = np.zeros(10) + 0.01

        targets[int(values[0])] = 0.99
        cnn.train(inputs, targets)
print('done!')

done!

with open('mnist_test.csv', 'r') as f:
    trains = f.readlines()
    alls = 0
    succe = 0
    for t in trains:
        values = t.split(",")
        inputs = (np.asfarray(values[1:]) / 255.0 * 0.99) + 0.01
        out = (cnn.query(inputs))
        image_array = np.asfarray(values[1:]).reshape((28, 28))
        import matplotlib
        import matplotlib.pyplot as plt
        #plt.imshow(image_array, cmap='Greys', interpolation='None')
        if list(out).index(max(list(out))) == int(values[0]):
            succe += 1
        alls += 1
    print(succe/float(alls))

0.9517

准确率95%

# 用自己手写的图片识别
import cv2
img = cv2.imread('2.png', cv2.IMREAD_GRAYSCALE)
data = img.flatten()
inputs = 1 - ((np.asfarray(data) / 255.0 * 0.99) + 0.01)
out = (cnn.query(inputs))
image_array = (255 - np.asfarray(data)).reshape((28, 28))
plt.imshow(image_array, cmap='Greys', interpolation='None')
print(list(out).index(max(list(out))))

2

output_5_1.png