Python实现感知器

# -*- coding: utf-8 -*-
"""
Created on Fri Sep  1 11:56:12 2017

@author: piaodexin
"""
#
import pandas as pd
import numpy as np
#生成正标签数据
x1=np.random.randn(50,2)+40
y1=np.ones((50,1))
data1=np.hstack([x1,y1])
#生成负标签数据
x2=np.random.randn(50,2)
y2=-np.ones((50,1))
data2=np.hstack([x2,y2])
#将数据合并
data=np.vstack([data1,data2])

from sklearn import cross_validation
#将原始数据分成训练数据，和测试数据
train_x,test_x,train_y,test_y=cross_validation.train_test_split(data[:,:-1],data[:,-1],test_size=0.25,
                                                                random_state=0,stratify=data[:,-1])

#初始化权重和偏差
w=np.random.randn(2)
b=np.ones(1)
#设置学习率
a=0.1
c=0
#进行3000轮实验
while c<3000:
    for i in range(len(train_x)):
        #如果某次分类错误，则修改权值和偏差
        if train_y[i]*(np.sum(w*train_x[i])+b)<=0:
            w+=a*train_y[i]*train_x[i]
            b=a*train_y[i]
            
    c+=1
#最后在测试集上检验正确率
count=0
for i in range(len(test_x)):
    if train_y[i]*(np.sum(w*train_x[i])+b)<=0:
        count+=1
precision=(len(test_x)-count)/len(test_x)