Machine Learning Foundations q15

最新推荐文章于 2021-09-29 10:40:55 发布

原创最新推荐文章于 2021-09-29 10:40:55 发布 · 507 阅读

0 ·

CC 4.0 BY-SA版权

本文介绍了一种基于Python实现的简单感知机学习算法。通过从文件中读取数据集，利用numpy进行矩阵运算来更新权重向量，直至所有训练样本都能被正确分类。此过程涉及数据预处理、初始化权重向量及迭代更新等步骤。

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

import fileinput
import numpy as np


def install(fileName):
    xSet=[]
    ySet=[]
    for line in fileinput.input(fileName):
        num_str = line.split()
        xSet.append(map(float, num_str[0:-1]))
        ySet.append(int(num_str[-1]))
    return (np.matrix(xSet),np.matrix(ySet).T)

  
xSet,ySet=install('hw1_15_train.dat') #从训练集中读取
xSet = np.concatenate((np.ones((xSet.shape[0],1)), xSet), 1)

w = np.matrix(np.zeros(5)) #初始化w，[[ 0.  0.  0.  0.  0.]]
       
count = 0
while True:
    correct_num = 0      
    for i in xrange(np.shape(xSet)[0]): #从martix里提取行数
        xn = xSet[i]
        yn = ySet[i]
        dot = np.dot(xn,w.T)
        if dot*yn <=0:
            w += yn*xn
            count = count + 1
        else:
            correct_num = correct_num + 1
    if correct_num == 400:
        break
    else: 
        print correct_num
        print count, w