bp神经网络python代码

以下是一个简单的基于Python的BP神经网络的代码实现：

import numpy as np

# 定义sigmoid函数
def sigmoid(x):
    return 1 / (1 + np.exp(-x))

# 定义BP神经网络类
class BPNeuralNetwork:
    def __init__(self, input_nodes, hidden_nodes, output_nodes, learning_rate):
        self.input_nodes = input_nodes
        self.hidden_nodes = hidden_nodes
        self.output_nodes = output_nodes
        self.lr = learning_rate
        
        # 初始化权重矩阵
        self.weights_ih = np.random.normal(0.0, pow(self.hidden_nodes, -0.5), (self.hidden_nodes, self.input_nodes))
        self.weights_ho = np.random.normal(0.0, pow(self.output_nodes, -0.5), (self.output_nodes, self.hidden_nodes))
        
    def train(self, inputs_list, targets_list):
        inputs = np.array(inputs_list, ndmin=2).T
        targets = np.array(targets_list, ndmin=2).T
        
        # 前向传播
        hidden_inputs = np.dot(self.weights_ih, inputs)
        hidden_outputs = sigmoid(hidden_inputs)
        final_inputs = np.dot(self.weights_ho, hidden_outputs)
        final_outputs = sigmoid(final_inputs)
        
        # 计算误差
        output_errors = targets - final_outputs
        hidden_errors = np.dot(self.weights_ho.T, output_errors)
        
        # 反向传播
        self.weights_ho += self.lr * np.dot((output_errors * final_outputs * (1 - final_outputs)), np.transpose(hidden_outputs))
        self.weights_ih += self.lr * np.dot((hidden_errors * hidden_outputs * (1 - hidden_outputs)), np.transpose(inputs))
        
    def query(self, inputs_list):
        inputs = np.array(inputs_list, ndmin=2).T
        
        # 前向传播
        hidden_inputs = np.dot(self.weights_ih, inputs)
        hidden_outputs = sigmoid(hidden_inputs)
        final_inputs = np.dot(self.weights_ho, hidden_outputs)
        final_outputs = sigmoid(final_inputs)
        
        return final_outputs

使用方法：

# 创建一个BP神经网络对象
input_nodes = 784
hidden_nodes = 100
output_nodes = 10
learning_rate = 0.3
nn = BPNeuralNetwork(input_nodes, hidden_nodes, output_nodes, learning_rate)

# 训练神经网络
for record in training_data_list:
    inputs = (np.asfarray(record[1:]) / 255.0 * 0.99) + 0.01
    targets = np.zeros(output_nodes) + 0.01
    targets[int(record[0])] = 0.99
    nn.train(inputs, targets)

# 测试神经网络
scorecard = []
for record in test_data_list:
    correct_label = int(record[0])
    inputs = (np.asfarray(record[1:]) / 255.0 * 0.99) + 0.01
    outputs = nn.query(inputs)
    label = np.argmax(outputs)
    if label == correct_label:
        scorecard.append(1)
    else:
        scorecard.append(0)
        
scorecard_array = np.asarray(scorecard)
print("performance = ", scorecard_array.sum() / scorecard_array.size)