图文详解感知机算法原理及Python实现

class Perceptron:
    def __init__(self):
        self.w = np.mat([0,0])                                    # 初始化权重
        self.b = 0                                                  # 初始化偏置
        self.delta = 1                                              # 设置学习率为1
        self.train_set =  [[np.mat([3, 3]), 1], [np.mat([4, 3]), 1], [np.mat([1, 1]), -1]]  # 设置训练集
        self.history = []                                           # 训练历史

def update(self,error_point):
        self.w += self.delta*error_point[1]*error_point[0]
        self.b += self.delta*error_point[1]
        self.history.append([self.w.tolist()[0],self.b])

def judge(self,point):
        return point[1]*(self.w*point[0].T+self.b)

def train(self):       
     flag = True
     while(flag):
         count = 0
         for point in self.train_set:
             if(self.judge(point)<=0):
                 self.update(point)
             else:
                 count += 1
         if(count == len(self.train_set)):
             flag = False

def show():
        print("参数w,b更新过程：",perceptron.history)
        anim = animation.FuncAnimation(fig, animate, init_func=init, frames=len(perceptron.history), 
                                        interval=1000, repeat=False,blit=True)
        plt.show()