Python：使用cvxpy包實現SVM二分類（可以運行通）

 解決了別人代碼裏面的一個問題，

對y進行了np.reshape，完美運行！

'''
代碼可以運行的通
'''
import numpy as np
import cvxpy as cp
import matplotlib.pyplot as plt
from sklearn import datasets

# Problem data.

x ,y = datasets.make_blobs(n_samples=300, n_features=2,centers=2,cluster_std=[2.0,2.0],random_state=123)
y = y*2 - 1
y = np.reshape(y,(len(y),1))


m = x.shape[0]
n = x.shape[1]
C = 1

# Construct the problem.
w = cp.Variable((n,1))
b = cp.Variable()
xi = cp.Variable((m,1))

objective = cp.Minimize(0.5*cp.norm(w)**2+C*cp.sum(xi))
constraints = [cp.multiply(y, x*w + b) >= 1 - xi, xi >= 0]
prob = cp.Problem(objective, constraints)

# The optimal objective value is returned by `prob.solve()`.
# The optimal value for x is stored in `x.value`.
result = prob.solve()
print(w.value)
# The optimal Lagrange multiplier for a constraint is stored in
# `constraint.dual_value`.
print(prob.value)
#print(constraints[0].dual_value)
#print(0.5*np.linalg.norm(w.value, ord =2)**2+C*np.sum(xi.value))

xp = np.linspace(min(x[:,0]), max(x[:,0]), 100)
yp = - (w.value[0]*xp + b.value)/w.value[1]
yp1 = np.array(- (w.value[0]*xp + b.value - C)/w.value[1] )# margin boundary for support vectors for y=1
yp0 = np.array(- (w.value[0]*xp + b.value + C)/w.value[1] )# margin boundary for support vectors for y=0

idx0 = np.where(y==-1)
idx1 = np.where(y==1)

plt.plot(x[idx0, 0], x[idx0, 1], 'rx')
plt.plot(x[idx1, 0], x[idx1, 1], 'go')
plt.plot(xp, yp, '-b', xp, yp1, '--g', xp, yp0, '--r')
plt.title('decision boundary for a linear SVM classifier with C={}'.format(C) )
plt.show()