机器学习-神经网络

类别：机器学习个人学习笔记
算法：神经网络
数据：月亮数据集，Python自带
书籍：《机器学习实战》、周志华西瓜书《机器学习》、李航《统计学习方法》

import numpy as np
import numpy as np 
import warnings
from sklearn.datasets import load_digits 
from sklearn.metrics import confusion_matrix, classification_report 
from sklearn.preprocessing import LabelBinarizer 
import matplotlib.pyplot as plt

# sigmoid function
def logistic(x):  
    return 1/(1 + np.exp(-x))

# 对激活函数求导
def logistic_derivative(x):     
    return x*(1-x)

#作业1）定义对应的tanh(x)函数及在BP算法中对应的导数
def tanh(x):
    return (np.exp(x) - np.exp(-x))/(np.exp(x)+np.exp(-x))

def tanh_deriv(x):        # 这里的x带入的是tanh(x)
    return 1-x**2

class NeuralNetwork:   
    def __init__(self, layers, activation='logistic'):  
        """  
        :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):  #初始化层之间的权重
            self.weights.append(np.random.randn(layers[i - 1] , layers[i] ))   #gaussian分布生成随机矩阵 
        self.weights.append(np.random.randn(layers[i] , layers[i + 1])) #接上最后一层，输出无截距添加
            
            
    def fit(self, X, y, learning_rate=0.5, epochs=10000):         
        X = np.atleast_2d(X)   # X:样本点，特征至少是 二维的             
        y = np.array(y)
    
        for k in range(epochs):  #epchos代表迭代次数
            i = np.random.randint(X.shape[0])   #随机选取样本中的任意一个样本点
            a = [X[i]] #1*特征维数的向量
    
            for l in range(len(self.weights)):  #前馈神经网络，计算每个神经元的数值
                a.append(self.activation(np.dot(a[l], self.weights[l])))  #从l层计算l+1层神经元参数，a(l+1)=f(x(l)W(l)),其中f为激活函数
            error = y[i] - a[-1]  #计算最后一层的误差，即e(最后一层)
            deltas = [error * self.activation_deriv(a[-1])] #计算delta(最后一层)=e(最后一层)*h(a(最后一层)) h为对应激活函数的导数
            
            #开始BP算法
            for l in range(len(a) - 2, 0, -1): # 从倒数第二层开始计算delta 
                deltas.append(deltas[-1].dot(self.weights[l].T)*self.activation_deriv(a[l]))  #delta(l)=(delta(l+1)*W(l)的转置).*h(a(l))
            deltas.reverse()  #将delta按照前馈神经网络顺序排列
            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)  #每层更新的偏导数为a[i]的转秩*delta[i]
            print(self.weights)
                
                
    def predict(self, x):        #针对进来的样本数据通过训练好的网络进行预测 
        x = np.array(x)         
        a=x
        for l in range(0, len(self.weights)):             
            a = self.activation(np.dot(a, self.weights[l]))         
        return a

nn = NeuralNetwork([2,3,1],