def plotJwiththeta2(x, y):
m, n = shape(x) #m: number of training example; n: number of features
x = c_[ones(m), x] #add x0
x = mat(x) # to matrix
y = mat(y)
maxcycle = 90000
theta = zeros((n + 1, 1)) #initial theta
theta[1] = -50
J = []
theta1 = zeros((maxcycle, 1))
for i in range(maxcycle):
h = lgr.sigmoid(x * theta)
theta[1] = theta[1] + 0.003
theta1[i] = theta[1]
#直接用append会出错，没有中间变量地址
cost = lgr.costfunction(y, h)
J.append(cost)
# print theta[1]
fig = plt.figure()
plt.plot(theta1, J)
plt.xlabel('theta1')
plt.ylabel('J')
plt.show()
return
def stocGradAscent0(dataMatrix, classLabels):
m, n = shape(dataMatrix)
alpha = 0.5
weights = ones(n) #initialize to all ones
weightsHistory = zeros((500 * m, n))
for j in range(500):
for i in range(m):
h = logRegres.sigmoid(sum(dataMatrix[i] * weights))
error = classLabels[i] - h
weights = weights + alpha * error * dataMatrix[i]
weightsHistory[j * m + i, :] = weights
return weightsHistory
def stocGradAscent0(dataMatrix, classLabels):
m,n = shape(dataMatrix)
alpha = 0.5
weights = ones(n) #initialize to all ones
weightsHistory=zeros((500*m,n))
for j in range(500):
for i in range(m):
h = logRegres.sigmoid(sum(dataMatrix[i]*weights))
error = classLabels[i] - h
weights = weights + alpha * error * dataMatrix[i]
weightsHistory[j*m + i,:] = weights
return weightsHistory
def classifyVector(inX, weights):
"""
Logistic回归分类函数
:param inX:
:param weights:
:return:
"""
prob = logRegres.sigmoid(sum(inX * weights))
if prob > 0.5:
return 1.0
else:
return 0.0
def stocGradAscent1(dataMatrix, classLabels):
m, n = shape(dataMatrix)
alpha = 0.4
weights = ones(n) #initialize to all ones
weightsHistory = zeros((40 * m, n))
for j in range(40):
dataIndex = range(m)
for i in range(m):
alpha = 4 / (1.0 + j + i) + 0.01
randIndex = int(random.uniform(0, len(dataIndex)))
h = logRegres.sigmoid(sum(dataMatrix[randIndex] * weights))
error = classLabels[randIndex] - h
#print error
weights = weights + alpha * error * dataMatrix[randIndex]
weightsHistory[j * m + i, :] = weights
del (dataIndex[randIndex])
print(weights)
return weightsHistory
def stocGradAscent1(dataMatrix, classLabels):
m,n = shape(dataMatrix)
alpha = 0.4
weights = ones(n) #initialize to all ones
weightsHistory=zeros((40*m,n))
for j in range(40):
dataIndex = range(m)
for i in range(m):
alpha = 4/(1.0+j+i)+0.01
randIndex = int(random.uniform(0,len(dataIndex)))
h = logRegres.sigmoid(sum(dataMatrix[randIndex]*weights))
error = classLabels[randIndex] - h
#print error
weights = weights + alpha * error * dataMatrix[randIndex]
weightsHistory[j*m + i,:] = weights
del(dataIndex[randIndex])
print weights
return weightsHistory
def stoc_grad_ascent0(data_matrix, class_label):
"""
随机梯度
这个算法不太好 至少误判了三分之一的数据
:param data_matrix:
:param class_label:
:return:
"""
m, n = data_matrix.shape
# 步长
alpha = 0.01
# 初始化所有的回归系数为1
weights = np.ones(n)
for i in range(m):
# 先每一个样本与对应回归系数weights相乘 再求和 最后在带入sigmoid求值
h = logRegres.sigmoid(np.sum(data_matrix[i] * weights))
# 绝对误差
error = class_label[i] - h
# 修正误差
weights = weights + alpha * error * data_matrix[i]
return weights
def stoc_grad_ascent1(data_matrix, class_label, max_iter=150):
"""
改进版随机梯度
:param data_matrix:
:param class_label:
:return:
"""
m, n = data_matrix.shape
# 初始化所有的回归系数为1
weights = np.ones(n)
for j in range(max_iter):
data_index = list(range(m))
for i in range(m):
# 步长
alpha = 4 / (1.0 + j + i) + 0.01
rand_index = int(np.random.uniform(0, len(data_index)))
# 先每一个样本与对应回归系数weights相乘 再求和 最后在带入sigmoid求值
h = logRegres.sigmoid(np.sum(data_matrix[rand_index] * weights))
# 绝对误差
error = class_label[rand_index] - h
# 修正误差
weights = weights + alpha * error * data_matrix[rand_index]
del data_index[rand_index]
return weights
y = np.array([[1, 1, 0, 1]]).T
# seed random numbers to make calculation
# deterministic (just a good practice)
np.random.seed(1)
# initialize weights randomly with mean 0
syn0 = 2 * np.random.random((3, 1)) - 1
print 'X:', X
print 'y:', y
print 'syn0:', syn0
# matrix multiplicative
print 'dot:', np.dot(X, syn0)
print 'dot:', logRegres.sigmoid(np.dot(X, syn0))
for iterator in xrange(1000):
# forward propagation
l0 = X
l1 = nonlin(np.dot(l0, syn0))
# how much did we miss?
l1_error = y - l1
# multiply how much we missed by the
# slope of the sigmssoid at the values in l1
l1_delta = l1_error * nonlin(l1, True)
# update weights
syn0 += np.dot(l0.T, l1_delta)
def classifyVector(inX, weight):
prob = logRegres.sigmoid(sum(inX * weight))
if prob > 0.5:
return 1.0
else:
return 0.0
def classifyVector(inX, weights):
prob = logRegres.sigmoid(np.sum(inX * weights))
return 1.0 if prob > 0.5 else 0.0
def predict(inX, weights):
probability = sigmoid(sum(np.array(inX) * weights))
if probability > 0.5:
return 1
else:
return 0
def classifyVector(inX, weights):
prob = lr.sigmoid(sum(inX*weights))
if prob > 0.5:
return 1
else:
return 0
def classifyVector(inX, weights):
prob = lr.sigmoid(sum(inX * weights))
if prob > 0.5:
return 1
else:
return 0
# coding=utf-8
import numpy
import logRegres
dataMat, labelMat = logRegres.loadDataSet()
weights = logRegres.gradAscent(dataMat, labelMat)
print weights
#logRegres.plotBestFit(weights.getA())
weights = logRegres.stocGradAscent0(dataMat, labelMat)
print '--随机梯度'
#logRegres.plotBestFit(weights)
weights = logRegres.stocGradAscent1(dataMat, labelMat)
print '--改进的随机梯度'
logRegres.plotBestFit(weights)
import matplotlib.pyplot as plt
fig = plt.figure()
ax = fig.add_subplot(111)
x = numpy.arange(-8.0, 8.0, 0.2)
print logRegres.sigmoid(-0.3)
# 设置偏移量 不是所有数据都是x=0划分
y = [logRegres.sigmoid(xi - 0.1) for xi in x]
ax.plot(x, y)
plt.xlabel('X1')
plt.ylabel('X2')
#plt.show()
print '--预测病马'
logRegres.multiTest()