# coding=utf-8
from numpy import *
import matplotlib.pyplot as plt

# Sigmoid激活函数，这里也给了tanh激活函数
def sigmoid(X):
    return 1.0/(1+exp(-X))
    #return 2*1.0/(1+exp(-2*X))-1  # tanh(x):mean value is 0

# 用于显示后面梯度上升训练过程中的weights
def trainingProcessDisplay(weights):
    fig = plt.figure()
    n = len(weights)
    x = range(n)
    ax1 = fig.add_subplot(311)
    ax1.plot(x,weights[:,0])
    plt.ylabel('w0')
    ax = fig.add_subplot(312)
    ax.plot(x,weights[:,1])
    plt.ylabel('w1')
    ax = fig.add_subplot(313)
    ax.plot(x,weights[:,2])
    plt.ylabel('w2')
    plt.xlabel('Iterations')
    plt.show()

# 梯度上升，每次更新都是计算出所有样本之间的误差（数据量很大时计算很耗时）
def gradAscent(data,labels, alpha = 0.001, numIter=500):
    dataMatrix = mat(data)
    labelMat = mat(labels).transpose()
    m,n = shape(dataMatrix)
    weights = ones((n,1))
    for k in range(numIter):
        h = sigmoid(dataMatrix*weights)       # 激活输出
        error = labelMat - h                  # 误差计算，上图中的蓝色框公式
        weights = weights + alpha*dataMatrix.transpose()*error    # 权值更新，这里使用批量数据
    return array(weights)

# 梯度上升，每次使用单个样本进行更新，error为标量
def stocGradAscent0(data,labels, alpha = 0.001, numIter=500):
    m,n = shape(data)
    weights = ones(n)
    for j in range(numIter):
        for i in range(m):
            h = sigmoid(sum(data[i]*weights))
            error = labels[i]-h
            weights = weights + alpha*error*data[i]
    return weights

# 随机梯度上升，每次迭代过程中顺序不一样，同样也是单样本更新
def stocGradAscent1(data,labels, numIter=500):
    m,n = shape(data)
    weights = ones(n)
    allWeights = ones((numIter,n))
    for j in range(numIter):
        allWeights[j] = weights
        dataIndex =  range(m)
        random.shuffle(dataIndex)
        for i in range(m):
            alpha = 4/(1.0+i+j) + 0.0001
            index = dataIndex[i]
            h = sigmoid(sum(data[index]*weights))
            error = labels[index] - h
            weights = weights + alpha *error*data[index]
    trainingProcessDisplay(allWeights)
    return weights

# 结果可视化：在这里回归出来的是直线
def resultVisualization(data,labels,weights):
    n = shape(data)[0]
    x1 = [];y1=[];x2=[];y2=[]
    for i in range(n):
        if int(labels[i])==1:
            x1.append(data[i][1]);y1.append(data[i][2])
        else:
            x2.append(data[i][1]);y2.append(data[i][2])
    fig = plt.figure()
    ax = fig.add_subplot(111)
    ax.scatter(x1,y1,s=30,c='red',marker='s')
    ax.scatter(x2,y2,s=30,c='green')
    x = arange(-5.0,5.0,0.1)
    # div = weights[2]
    # if weights[2]==0: div = 0.00001
    y = (-weights[0]-weights[1]*x)/weights[2]
    ax.plot(x,y)
    plt.xlabel('X1');plt.ylabel('X2')
    plt.show()

# 测试
def testLR():
    data,labels = loadData("data.txt")
    dataArr = array(data)
    #weights = gradAscent(dataArr,labels)
    #weights = stocGradAscent0(dataArr,labels)
    weights = stocGradAscent1(dataArr,labels)
    #print weights
    resultVisualization(dataArr,labels,weights)

if __name__ == "__main__":
    testLR()