【基于python实现k-means聚类算法-矩阵运算】

基于python实现k-means聚类算法-矩阵运算

说明：本文仅仅用代码实现自己对Kmeans聚类算法理解。本次实现采用矩阵计算的方式实现kmeans聚类算法，算是对基于python实现k-means聚类算法的补充。

import numpy as np
import matplotlib.pyplot as plt


# # 计算Euclidean distance
def eucdistance(v1, v2):
    # if v1.ndim == 1:
    #     v1 = v1.reshape(1, -1)
    v1 = v1.reshape(1, -1) if v1.ndim == 1 else v1
    v2 = (v2.reshape(1, -1) if v2.ndim == 1 else v2)
    dist = np.sqrt(np.sum(np.power(v1 - v2, 2), -1))
    return dist


def init_k_centers(x, k):
    assert x.ndim == 2, "x's dim not is 2"
    k_centers = x[0: k]
    return k_centers


def k_means_v2(x_data, k=2, ITERS=20, showfig=True, colors=["g", "y", "b"]):
    n, f_n = x_data.shape
    last_labels = 0
    results = {}

    for iter in range(ITERS):
        init_centers = init_k_centers(x_data, k=k)
        data_kn = x_data[None, :, :].repeat(repeats=k, axis=0)  # (n, f_n) -> (K, n, f_n)
        kn_centers = init_centers[:, None, :].repeat(repeats=n, axis=1)  # (k, f_n) -> (k, n, f_n)

        dist = np.sqrt(np.power(data_kn - kn_centers, 2).sum(axis=2))
        labels = dist.argmin(axis=0)

        # 更新聚类中心
        for i in range(k):
            init_centers[i] = x_data[labels == i].mean(axis=0).reshape(-1, f_n)
            x_data_k = x_data[np.array(labels) == i]
            results[i] = x_data_k

        if np.sum(labels != last_labels) == 0 or (iter == ITERS - 1):
            return results

        last_labels = labels

        if showfig:
            plt.ion()
            plt.show()
            plt.cla()

            for key in set(labels):
                k_samples = x_data[labels == key]
                plt.scatter(k_samples[:, 0], k_samples[:, 1], color=colors[key], marker=".", s=8)
                plt.scatter(init_centers[key][0], init_centers[key][1], color=colors[key], marker="*", s=15)
                plt.text(1.5, 0.5, "iter=%.2i" % (iter))
                plt.pause(0.1)

            plt.ioff()
            plt.show()


if __name__ == '__main__':
    # x = np.random.randn(100, 2)
    x_data = np.array([[3, 10], [2, 9], [1, 9], [3, 7], [4, 8], [3.5, 6], [9, 0.5],
                       [8, 3], [9, 2], [8, 1], [10, 0.5], [6.5, 2], [6.5, 4]])

    # results = k_means_v1(x_data, k=2, ITERS=20, error_thred=1e-3, random_state=42, showfig=True, colors=["g", "b"])

    result = k_means_v2(x_data, k=2, ITERS=20, showfig=True, colors=["g", "y", "b"])

    print("*." * 20 + " k-means done " + "*." * 20)