参考李航《统计学习方法》 一开始的感知机章节，看着不太复杂就实现一下。。。
"""
感知机学习算法的原始形式
例2.1
"""
import numpy as np class Perceptron:
def __init__(self,w,b,alpha):
self.w = w
self.b = b
self.alpha = alpha def loss(self,x,y):
return np.sum( y*(np.dot(x, self.w) + self.b) ) def sgd(self,x,y): # 随机梯度下降函数
self.w += self.alpha * y * x
self.b += self.alpha * y def train(self,X,Y):
while(True):
M = len(X) # 错误分类数
for i in range(len(X)):
if self.loss(X[i],Y[i])<=0:
self.sgd(X[i],Y[i])
print "w:",self.w," b:",self.b
else:
M -= 1
if not M:
print "final optimal:","w:",self.w," b:",self.b
break class Perceptron_dual:
def __init__(self,alpha,b,ita):
self.alpha = alpha
self.b = b
self.ita = ita def gram(self,X):
return np.dot(X,X.T) def train(self,X,Y):
g = self.gram(X) M = len(X) # 错误分类数
while(True):
M = len(X) # 错误分类数
for j in range(len(X)):
if Y[j] * (np.sum(self.alpha * Y * g[j]) + self.b) <= 0:
self.alpha[j] += self.ita
self.b += self.ita * Y[j]
print "a:",self.alpha," b:",self.b
else:
M -= 1
if M == 0:
print "final optimal:","a:",self.alpha," b:",self.b
break if __name__ == "__main__": X = np.array([[3,3],[4,3],[1,1]]) Y = np.array([1,1,-1])
perc_d = Perceptron_dual(np.zeros(Y.shape),0,1)
perc_d.train(X, Y)