CV学习笔记-BP神经网络代码Python实例

一. 任务描述

给定数据集（txt文件），采用随机梯度下降的方式进行神经网络的学习，完成权重参数的更新，使得输入的数据能够接近输出label。

关于BP神经网络的手推和原理见笔者《CV学习笔记-推理和训练》、《CV学习笔记-BP神经网络》

txt文件类似下图所示

二. 程序设计

1. 神经网络设计

类：NeuralNetWork
类内初始化： __init__用以设置神经网络的参数（输入层参数、隐藏层参数、输出层参数、学习率）
类内方法： train用于训练数据，更新权重
读取数据： loadDataSet用于在txt文件中读取数据，包括输入值和label值
随机梯度下降处理： stocGradDescent用于处理训练数据的过程

2. 具体设计

• NeuralNetWork类

class NeuralNetWork:
    def __init__(self, input_nodes, hidden_nodes, out_nodes, lr):
    	# 设置输入个数
        self.innodes = input_nodes
        # 设置隐藏层节点个数
        self.hnodes = hidden_nodes
        # 设置输出节点个数
        self.onodes = out_nodes
        # 设置学习率，用于反向更新
        self.lr = lr
        # self.weight_i2h = np.ones((self.hnodes, self.innodes))
        # self.weight_h2o = np.ones((self.onodes, self.hnodes))
        # 随机初始化比1矩阵效果要好很多
        # 权重矩阵（输入到隐藏）
        self.weight_i2h = (numpy.random.normal(0.0, pow(self.hnodes,-0.5), (self.hnodes,self.innodes) )  )
        # 权重矩阵（隐藏到输出）
        self.weight_h2o = (numpy.random.normal(0.0, pow(self.onodes,-0.5), (self.onodes,self.hnodes) )  )
        # 设置激活函数（sigmoid）
        self.activation_function = lambda x: 1.0/(1+np.exp(-x))
        pass

	'''
	训练方法，输入一次训练的输入和label
	'''
    def train(self, inputs_list, targets_list):
        inputs = numpy.array(inputs_list, ndmin=2).T
        target = np.array(targets_list, ndmin=2).T
        # wx+b
        hidden_inputs = np.dot(self.weight_i2h, inputs)
        # 激活作为隐藏层的输出
        hidden_outputs = self.activation_function(hidden_inputs)
        # wx+b
        o_inputs = np.dot(self.weight_h2o, hidden_outputs)
        # 激活作为输出
        o_outputs = self.activation_function(o_inputs)
        # 损失函数
        loss = (target - o_outputs) ** 2 * 0.5
        # 输出误差，用于反向更新
        error = target - o_outputs
        # error = target - o_outputs
        # 隐藏层误差，用于反向更新
        hidden_error = np.dot(self.weight_h2o.T, error * o_outputs * (1 - o_outputs))
        # 梯度
        gradO = error * o_outputs * (1 - o_outputs)
        # 反向更新，详见笔者博客[《CV学习笔记-BP神经网络》(https://blog.csdn.net/qq_38853759/article/details/121930413)
        self.weight_h2o += self.lr * np.dot((error * o_outputs * (1 - o_outputs)), np.transpose(hidden_outputs))
        gradI = hidden_error * hidden_outputs * (1 - hidden_outputs)
        # 反向更新
        self.weight_i2h += self.lr * np.dot((hidden_error * hidden_outputs * (1 - hidden_outputs)),
                                            np.transpose(inputs))
        return loss

• 读取数据集并处理

def loadDataSet():
    data = []
    label = []
    fr = open('testSet.txt')
    for line in fr.readlines():
        lineArr = line.strip().split()
        data.append([1.0, float(lineArr[0]), float(lineArr[1])])
        label.append(int(lineArr[2]))
    data = np.array(data)
    label = np.array(label)
    return data, label

• 训练过程处理

def stocGradDescent(data, label):
    m, n = np.shape(data)
    for iter in range(200):
        total_loss = 0
        for i in range(m):
            # if label[i] == 1:
            #     pass
            # elif label[i] == 0:
            #     pass
            # 累计每个epoch的loss观察效果
            total_loss += net.train(data[i], label[i])
        print("NO.{} Loss={}".format(iter, total_loss))

三、实践代码

import numpy
import numpy as np


class NeuralNetWork:
    def __init__(self, input_nodes, hidden_nodes, out_nodes, lr):
        self.innodes = input_nodes
        self.hnodes = hidden_nodes
        self.onodes = out_nodes
        self.lr = lr
        # self.weight_i2h = np.ones((self.hnodes, self.innodes))
        # self.weight_h2o = np.ones((self.onodes, self.hnodes))
        # 随机初始化比1矩阵效果要好很多
        self.weight_i2h = (numpy.random.normal(0.0, pow(self.hnodes,-0.5), (self.hnodes,self.innodes) )  )
        self.weight_h2o = (numpy.random.normal(0.0, pow(self.onodes,-0.5), (self.onodes,self.hnodes) )  )
        self.activation_function = lambda x: 1.0/(1+np.exp(-x))
        pass

    def train(self, inputs_list, targets_list):
        inputs = numpy.array(inputs_list, ndmin=2).T
        target = np.array(targets_list, ndmin=2).T
        hidden_inputs = np.dot(self.weight_i2h, inputs)
        hidden_outputs = self.activation_function(hidden_inputs)
        o_inputs = np.dot(self.weight_h2o, hidden_outputs)
        o_outputs = self.activation_function(o_inputs)
        loss = (target - o_outputs) ** 2 * 0.5
        error = target - o_outputs
        # error = target - o_outputs
        hidden_error = np.dot(self.weight_h2o.T, error * o_outputs * (1 - o_outputs))
        gradO = error * o_outputs * (1 - o_outputs)
        self.weight_h2o += self.lr * np.dot((error * o_outputs * (1 - o_outputs)), np.transpose(hidden_outputs))
        gradI = hidden_error * hidden_outputs * (1 - hidden_outputs)
        self.weight_i2h += self.lr * np.dot((hidden_error * hidden_outputs * (1 - hidden_outputs)),
                                            np.transpose(inputs))
        return loss


# 从testSet.txt中读取数据存储至样本集data和标签集label
def loadDataSet():
    data = []
    label = []
    fr = open('testSet.txt')
    for line in fr.readlines():
        lineArr = line.strip().split()
        data.append([1.0, float(lineArr[0]), float(lineArr[1])])
        label.append(int(lineArr[2]))
    data = np.array(data)
    label = np.array(label)
    return data, label


def stocGradDescent(data, label):
    m, n = np.shape(data)
    for iter in range(200):
        total_loss = 0
        for i in range(m):
            # if label[i] == 1:
            #     pass
            # elif label[i] == 0:
            #     pass
            total_loss += net.train(data[i], label[i])
        print("NO.{} Loss={}".format(iter, total_loss))


if __name__ == '__main__':
    input_nodes = 3
    hidden_nodes = 3
    output_nodes = 1
    learning_rate = 0.1
    net = NeuralNetWork(input_nodes, hidden_nodes, output_nodes, learning_rate)
    data, label = loadDataSet()
    stocGradDescent(data, label)
    print(net.weight_i2h)
    print(net.weight_h2o)
    pass

txt文本文件内容请自行复制

-0.017612	14.053064	0
-1.395634	4.662541	1
-0.752157	6.538620	0
-1.322371	7.152853	0
0.423363	11.054677	0
0.406704	7.067335	1
0.667394	12.741452	0
-2.460150	6.866805	1
0.569411	9.548755	0
-0.026632	10.427743	0
0.850433	6.920334	1
1.347183	13.175500	0
1.176813	3.167020	1
-1.781871	9.097953	0
-0.566606	5.749003	1
0.931635	1.589505	1
-0.024205	6.151823	1
-0.036453	2.690988	1
-0.196949	0.444165	1
1.014459	5.754399	1
1.985298	3.230619	1
-1.693453	-0.557540	1
-0.576525	11.778922	0
-0.346811	-1.678730	1
-2.124484	2.672471	1
1.217916	9.597015	0
-0.733928	9.098687	0
-3.642001	-1.618087	1
0.315985	3.523953	1
1.416614	9.619232	0
-0.386323	3.989286	1
0.556921	8.294984	1
1.224863	11.587360	0
-1.347803	-2.406051	1
1.196604	4.951851	1
0.275221	9.543647	0
0.470575	9.332488	0
-1.889567	9.542662	0
-1.527893	12.150579	0
-1.185247	11.309318	0
-0.445678	3.297303	1
1.042222	6.105155	1
-0.618787	10.320986	0
1.152083	0.548467	1
0.828534	2.676045	1
-1.237728	10.549033	0
-0.683565	-2.166125	1
0.229456	5.921938	1
-0.959885	11.555336	0
0.492911	10.993324	0
0.184992	8.721488	0
-0.355715	10.325976	0
-0.397822	8.058397	0
0.824839	13.730343	0
1.507278	5.027866	1
0.099671	6.835839	1
-0.344008	10.717485	0
1.785928	7.718645	1
-0.918801	11.560217	0
-0.364009	4.747300	1
-0.841722	4.119083	1
0.490426	1.960539	1
-0.007194	9.075792	0
0.356107	12.447863	0
0.342578	12.281162	0
-0.810823	-1.466018	1
2.530777	6.476801	1
1.296683	11.607559	0
0.475487	12.040035	0
-0.783277	11.009725	0
0.074798	11.023650	0
-1.337472	0.468339	1
-0.102781	13.763651	0
-0.147324	2.874846	1
0.518389	9.887035	0
1.015399	7.571882	0
-1.658086	-0.027255	1
1.319944	2.171228	1
2.056216	5.019981	1
-0.851633	4.375691	1
-1.510047	6.061992	0
-1.076637	-3.181888	1
1.821096	10.283990	0
3.010150	8.401766	1
-1.099458	1.688274	1
-0.834872	-1.733869	1
-0.846637	3.849075	1
1.400102	12.628781	0
1.752842	5.468166	1
0.078557	0.059736	1
0.089392	-0.715300	1
1.825662	12.693808	0
0.197445	9.744638	0
0.126117	0.922311	1
-0.679797	1.220530	1
0.677983	2.556666	1
0.761349	10.693862	0
-2.168791	0.143632	1
1.388610	9.341997	0
0.317029	14.739025	0

四、 效果测试

通过debug看效果：

可见刚开始网络的输出跟label的差距还是很大

可以看到训练到一百多代的时候loss已经有了明显的下降

此时debug看效果，可以看到效果明显改善，网络的输出已经比较小接近于真实label：0了

label为1的效果也靠谱很多了

最后打印出权重参数：