感知机学习算法 python实现
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。。。
1 """ 2 感知机学习算法的原始形式 3 例2.1 4 """ 5 import numpy as np 6 7 class Perceptron: 8 def __init__(self,w,b,alpha): 9 self.w = w 10 self.b = b 11 self.alpha = alpha 12 13 def loss(self,x,y): 14 return np.sum( y*(np.dot(x, self.w) + self.b) ) 15 16 def sgd(self,x,y): # 随机梯度下降函数 17 self.w += self.alpha * y * x 18 self.b += self.alpha * y 19 20 def train(self,X,Y): 21 while(True): 22 M = len(X) # 错误分类数 23 for i in range(len(X)): 24 if self.loss(X[i],Y[i])<=0: 25 self.sgd(X[i],Y[i]) 26 print "w:",self.w," b:",self.b 27 else: 28 M -= 1 29 if not M: 30 print "final optimal:","w:",self.w," b:",self.b 31 break 32 33 class Perceptron_dual: 34 def __init__(self,alpha,b,ita): 35 self.alpha = alpha 36 self.b = b 37 self.ita = ita 38 39 def gram(self,X): 40 return np.dot(X,X.T) 41 42 def train(self,X,Y): 43 g = self.gram(X) 44 45 M = len(X) # 错误分类数 46 while(True): 47 M = len(X) # 错误分类数 48 for j in range(len(X)): 49 if Y[j] * (np.sum(self.alpha * Y * g[j]) + self.b) <= 0: 50 self.alpha[j] += self.ita 51 self.b += self.ita * Y[j] 52 print "a:",self.alpha," b:",self.b 53 else: 54 M -= 1 55 if M == 0: 56 print "final optimal:","a:",self.alpha," b:",self.b 57 break 58 59 if __name__ == "__main__": 60 61 X = np.array([[3,3],[4,3],[1,1]]) 62 63 Y = np.array([1,1,-1]) 64 perc_d = Perceptron_dual(np.zeros(Y.shape),0,1) 65 perc_d.train(X, Y)
