统计机器学习笔记(一)---感知机模型

代码

一、公共绘图部分

import numpy as np
from matplotlib import pyplot as plt
from matplotlib import animation
from functools import partial

# first set up the figure, the axis, and the plot element we want to animate
__FIG = plt.figure()
__AX = plt.axes(xlim=(0, 2), ylim=(-2, 2))
__LINE, = __AX.plot([], [], 'g', lw = 2)
__LABEL = __AX.text([], [], '')

# initialization function: plot the background of each frame
def __init(trains):
    __LINE.set_data([], [])
    x, y, x_, y_ = [], [], [], []
    for p in trains:
        if p[1] > 0:
            x.append(p[0][0])
            y.append(p[0][1])
        else:
            x_.append(p[0][0])
            y_.append(p[0][1])

    plt.plot(x, y, 'bo', x_, y_, 'rx')
    plt.axis([-6, 6, -6, 6])
    plt.grid(True)
    plt.xlabel('x')
    plt.ylabel('y')
    plt.title('Prceptron Algorithm(wendale.cn)')
    return __LINE, __LABEL

# animation function.  this is called sequentially
def __animate(history, i):
    global __AX, __LINE, __LABEL

    w = history[i][0]
    b = history[i][1]
    if w[1] == 0:
        return __LINE, __LABEL
    x1 = -7
    y1 = -(b + w[0]*x1)/w[1]
    x2 = 7
    y2 = -(b + w[0]*x2)/w[1]
    __LINE.set_data([x1, x2], [y1, y2])
    x1 = 0
    y1 = -(b + w[0] * x1) / w[1]
    __LABEL.set_text(history[i])
    __LABEL.set_position([x1, y1])
    return __LINE, __LABEL

def show(trains, history, save=False, name=''):
    # call the animator.  blit=true means only re-draw the parts that have changed.
    anim = animation.FuncAnimation(__FIG, partial(__animate, history), init_func=partial(__init, trains),
                                   frames=len(history), interval=1000, repeat=True, blit=True)
    plt.show()
    if save:
        anim.save(name, fps=2, writer='imagemagick')

if __name__ == '__main__':
    trains = np.array([[(3, 3), 1], [(4, 3), 1], [(1, 1), -1]])
    history = [[[1.0, 0.0, -0.0], 1], [[1.0, 0.0, -1.0], 0], [[1.0, 0.0, -2.0], -1], [[1.0, 0.0, -3.0], -2], [[2.0, 0.0, -3.0], -1], [[2.0, 0.0, -4.0], -2], [[2.0, 0.0, -5.0], -3]]
    show(trains, history)

二、原始形式

import copy
import draw

HISTORY = []
TRAINS = [[(3, 3), 1], [(4, 3), 1], [(1, 1), -1]]

def update(w, b, item):
    for i in range(len(item[0])):
        w[i] += item[1] * item[0][i]
    b += item[1]
    return w, b

def cal(item):
    res = 0
    for i in range(len(item[0])):
        res += item[0][i] * w[i]
    res += b
    res *= item[1]
    return res

def iterComp(w, b, trains):
    global HISTORY     
    flag = False
    for item in trains:
        if cal(item) <= 0:
            flag = True
            (w, b) = update(w, b, item)
            HISTORY.append([copy.copy(w), b])
    if not flag:
        print "RESULT:w:" + str(w) + "b:" + str(b)
    return flag, w, b

if __name__ == "__main__":
    w = [0, 0]
    b = 0   
    while True:
        (flag, w, b) = iterComp(w, b, TRAINS)
        if not flag:
            break
    draw.show(TRAINS, HISTORY, True, 'perceptron.gif')

效果

三、对偶形式

__author__ = 'wendale'

import numpy as np
import draw

def __cal_gram(trains):
    gram = np.empty((len(trains), len(trains)), np.int)
    for i in range(len(trains)):
        for j in range(len(trains)):
            gram[i][j] = np.dot(trains[i][0], trains[j][0])
    return gram

def __check(trains, gram, alpha, b):
    y = np.array(trains[:,1])
    for i in range(len(trains)):
        if (np.dot(alpha * y, gram[i]) + b) * y[i] <= 0:
            return False, i
    return True, -1

def find(trains):
    gram = __cal_gram(trains)
    alpha = np.zeros(len(trains))
    b = 0
    x = np.array(trains[:,0].tolist()).reshape(3,2)
    y = np.array(trains[:,1])
    history = []
    while True:
        (flag, i) = __check(trains, gram, alpha, b)
        if flag:
            break
        #update
        alpha[i] += 1
        b += trains[i][1]
        history.append([np.dot(alpha * y, x).tolist(), b])
    return alpha, b, history

if __name__ == '__main__':
    trains = np.array([[[3, 3], 1], [[4, 3], 1], [[1, 1], -1]])
    (a, b, history) = find(trains)
    draw.show(trains, history, True, 'duality_perceptron.gif')

效果

参考

参考1