统计学习方法 第6章 习题

6.1确认逻辑斯蒂分布属于指数分布族：确认

6.2逻辑斯蒂回归模型的梯度下降法

# -*- coding: utf-8 -*-


import numpy as np
import matplotlib.pyplot as plt

            
class Logistic():
    def __init__(self,X,Y):
        self.X = np.c_[X,np.ones(X.shape[0])]
        self.Y = Y
        self.w = np.zeros(self.X.shape[1])
        
    def prob(self,w,xin):
        prb = 1.0/(1+np.exp(-np.dot(w,xin)))
        return 1-prb,prb
    
    def training(self,e,rate):
        def Lw(X,Y,w):
            sum = 0.0
            for i,yi in enumerate(Y):
                sum += yi*np.dot(w,X[i])-np.log(1+np.exp(np.dot(w,X[i])))
            return sum
                
        def gradient(X,Y,w):
            sum = np.zeros(X.shape[1])
            for i,xi in enumerate(X):
                _,prb = self.prob(w,xi)
                sum += xi*(Y[i] - prb)
            return sum  
        
        #附录A 梯度下降
        #第一步初始化
        X = self.X
        Y = self.Y
        w = self.w
        prelw = 0
        gw = 0
        gwf = float("inf")#直接进入迭代
        #当梯度范数小于e时停止迭代
        while gwf > e :#可以加上最大迭代次数
            w += rate*gw
            #计算L（w）
            nowlw = Lw(X, Y, w)
            print(w)
            #当前后的似然函数变化小于e时停止迭代
            if abs(prelw-nowlw) < e:
                break
            else:
                prelw = nowlw
                #计算梯度
                gw = gradient(X, Y, w)
                #梯度范数
                gwf = np.linalg.norm(gw,ord=2)                
                
        print("success")
        self.w = w
        print(w)
                
            

def main():
    np.random.seed(12)
    num_observations = 5000
    #这是数据集，这是借鉴的
    x1 = np.random.multivariate_normal([0, 0], [[1, .75],[.75, 1]], num_observations)
    x2 = np.random.multivariate_normal([1, 4], [[1, .75],[.75, 1]], num_observations)

    
    X &