Question

这是一堂课，非常感谢您的帮助！我根据收到的评论进行了一些更改，但现在又出现另一个错误。我需要修改一个实现均值平移算法的现有函数，但是该函数不会将所有点初始化为第一组质心，而是会根据半径基于栅格创建质心网格。我还需要删除不包含任何数据点的质心。我的问题是我不明白如何解决我收到的错误！

---------------------------------------------------------------------------
IndexError                                Traceback (most recent call last)
<ipython-input-7-de18ffed728f> in <module>()
     49 centroids = initialize_centroids(x)
     50 
---> 51 new_centroids = update_centroids(x, centroids, r = 1)
     52 
     53 print(len(centroids))
<ipython-input-7-de18ffed728f> in update_centroids(data, centroids, r)
     26         #print(len(centroids))
     27         #print(range(len(centroids)))
---> 28         centroid = centroids[i]
     29         for data_point in data:
     30             if np.linalg.norm(data_point - centroid) < r:
IndexError: index 2 is out of bounds for axis 0 with size 2

我尝试使用输入数据集的范围作为网格的边界，点之间用半径分隔。

from sklearn import datasets
import numpy as np
import matplotlib.pyplot as plt


def initialize_centroids(data, r = 1):
    '''Creates a grid of centroids with grid based on radius'''

    data = np.array(data)
    xi,yi = min(range(len(data))), max(range(len(data)))
    mx = np.arange(xi,yi,r)
    x,y = np.meshgrid(mx,mx)
    centroids=np.vstack([x.ravel(), y.ravel()])
    return centroids

#update centroids based on mean of points that fall within a specified radius of each centroid

def update_centroids(data, centroids, r = 1):

    new_centroids = []

    for i in centroids:
        in_radius = []
        centroid = centroids[i] #this is where the error occurs
        for data_point in data:
            if np.linalg.norm(data_point - centroid) < radius:
                in_radius.append(data_point) #this list is appended by adding the new centroid to it if the above conition is satisfied. 

        new_centroid = np.mean(in_radius, axis=0) 
        #maybe another way to do the next part
        new_centroids.append(tuple(new_centroid)) 

    unique_centroids = sorted(list(set(new_centroids))) #for element in in_radius, if element in set skip else set.append(element(in_rad)). append does not work with set. 

    new_centroids = {i:np.array(unique_centroids[i]) for i in range(len(unique_centroids))}

    return new_centroids

#test function on:
x, y = datasets.make_blobs(n_samples=300, n_features = 2, centers=[[0, 7], [0, -7], [5,7], [5, 0]])

centroids = initialize_centroids(x)

new_centroids = update_centroids(x, centroids, radius = 2)

print(len(centroids))
print()
print(len(new_centroids))

#code for plotting initially: 
plt.scatter(x[:,0], x[:,1], color = 'k')

for i in range(len(new_centroids)):
    plt.scatter(new_centroids[i][0], new_centroids[i][1], s=200, color = 'r', marker = "*")

#code for plotting updated centroids:
new_centroids = update_centroids(x, new_centroids, radius = 2)

plt.scatter(x[:,0], x[:,1], color = 'k')

for i in range(len(new_centroids)):
    plt.scatter(new_centroids[i][0], new_centroids[i][1], s=200, color = 'r', marker = "*")

#code for iterations:
def iterate_to_conv(data, max_iter=100):

    centroids = initialize_centroids(data)
    iter_count = 0

    while iter_count <= max_iter:
        new_centroids = update_centroids(data, centroids, radius = 2)
        centroids = new_centroids
        iter_count += 1


    return centroids


centroids = iterate_to_conv(x)

plt.scatter(x[:,0], x[:,1], color = 'k')

for i in range(len(centroids)):
    plt.scatter(centroids[i][0], centroids[i][1], s=200, color = 'r', marker = "*")

该函数需要返回最终质心的数量。我还没有走得很远，不知道整个均值平移实现将如何与该功能一起使用。.

Answer 1

在运行该循环时：for i in centroids通过质心迭代的i不是数字，而是一个向量，这就是为什么会弹出错误的原因。例如，第一个i值可能等于[0 1 2 0 1 2 0 1 2]。因此，将其作为索引是没有意义的。您的代码说的是采用centroid = centroid [n1 n2 nk]。要解决此问题，您确实需要更改初始化质心函数的工作方式。 Meshgrid也不会创建N维网格，因此您的Meshgrid可能适用于2维，但不适用于N维。希望对您有所帮助。

如何使用质心网格实现均值漂移？

1 个答案: