算法

 

 

代码

import numpy as np

def perceptron_training(X, y, learning_rate, max_iter_nums):
    iter_nums = 0
    [N, n] = X.shape
    w, b= np.zeros(n), 0
    while (iter_nums < max_iter_nums):
        no_mistake = True
        for i in range(N):
            if y[i] * (np.dot(w, X[i]) + b) <= 0:
                w += learning_rate * y[i] * X[i]
                b += learning_rate * y[i]
                no_mistake = False
        if no_mistake == True:
            break
        ++iter_nums
    return w, b

def test(X, y, w, b):
    N = len(X)
    no_mistake = True
    for i in range(N):
        if y[i] * (np.dot(w, X[i]) + b) <= 0:
            return False
    return True


# test: positive point: x1 = (3, 3), x2 = (4, 3); negative point: x3 = (1, 1)
X_test = np.array([[3, 3], [4, 3], [1, 1]])
y_test = np.array([1, 1, -1])
w, b = perceptron_training(X_test, y_test, 1, 30)
print("w: ", w)
print("b: ", b)
print("after traning: ", test(X_test, y_test, w, b))

测试结果