如何 select 将所有元素作为头部包含低于损失阈值的值

Question

我编写了一些代码用于对元素列表进行聚类，然后 select 每组中的一个元素具有最小损失值。

import warnings
import math
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns
from scipy.spatial import distance
from sklearn.metrics.pairwise import euclidean_distances
from sklearn.cluster import KMeans
from prettytable import PrettyTable
from tabulate import tabulate
%matplotlib inline
warnings.filterwarnings('ignore')

loss_data = np.array([1.453158736228, 4.71358454227, 2.22189044952, 1.369125723838, 1.294319868087, 3.604614257, 1.282392263412,
1.035766720771, 3.537582755088, 1.647409796714, 2.12351524829, 1.522501230239, 1.092259526252, 3.8876036, 2.12351319])

df=loss_data.reshape(-1, 1)

然后我根据元素的损失值对元素进行聚类

#calculate the clusters, labels and centres
KMeans = KMeans(n_clusters=3, random_state=0).fit(df)
clust_labels = KMeans.labels_.tolist() #labels of the elements classified to the cluster
clusters = KMeans.cluster_centers_ #cluster centres
print('clusters', clusters)
print('clust_labels', clust_labels)
loss = loss_data #store the losses
labels = clust_labels

之后为所有元素及其组和损失定义一个框架

#define clusters dataframe table
df_clusters = pd.DataFrame(index = range(len(df)))
df_clusters['element'] = list(df_clusters.index)

#display the clusters table
df_clusters['Clusters'] = clust_labels
df_clusters['loss'] = loss
df_clusters = df_clusters[['element', 'Clusters', 'loss']]

print ('Tabel 1: element with Clusters')
print(tabulate(df_clusters, headers='keys', tablefmt='psql'))

Tabel 1: element with their Clusters
+----+-----------+------------+---------+
|    |   element |   Clusters |    loss |
|----+-----------+------------+---------|
|  0 |         0 |          0 | 1.45316 |
|  1 |         1 |          1 | 4.71358 |
|  2 |         2 |          2 | 2.22189 |
|  3 |         3 |          0 | 1.36913 |
|  4 |         4 |          0 | 1.29432 |
|  5 |         5 |          1 | 3.60461 |
|  6 |         6 |          0 | 1.28239 |
|  7 |         7 |          0 | 1.03577 |
|  8 |         8 |          1 | 3.53758 |
|  9 |         9 |          0 | 1.64741 |
| 10 |        10 |          2 | 2.12352 |
| 11 |        11 |          0 | 1.5225  |
| 12 |        12 |          0 | 1.09226 |
| 13 |        13 |          1 | 3.8876  |
| 14 |        14 |          2 | 2.12351 |
+----+-----------+------------+---------+

然后我 select 一个具有最小值损失的元素

#define cluster heads dataframe table with min loss
df_clusters_heads = pd.DataFrame(index = range(3))
df_clusters_heads.index = df_clusters_heads.index
df_clusters_heads['Clusters'] = df_clusters_heads.index

min_loss = [] #list to store the min losses
clust_heads = [] #list to store cluster heads
for j in range(len(clusters)):
    elems = [i for i, x in enumerate(clust_labels) if x == j] #get elements of cluster k
    min_losses = [[loss[elem]] for elem in elems] #get losses for the element
    min_loss_temp = min(min_losses) #get the min loss
    min_loss.append(min_loss_temp) #store the min loss
    min_ind = min_losses.index(min_loss_temp) #get the element with min loss
    clust_heads.append(elems[min_ind]) #store the min loss element as cluster head

#display table for cluster heads and cluster
df_clusters_heads['Cluster Heads'] = clust_heads
df_clusters_heads["loss"] = min_loss
df_clusters_heads = df_clusters_heads[['Clusters', 'Cluster Heads', 'loss']]
        
print ('Tabel 2: heads of each  Cluster')
print(tabulate(df_clusters_heads, headers='keys', tablefmt='psql'))
print("")

Tabel 2: heads of each  Cluster
+----+------------+-----------------+------------------+
|    |   Clusters |   Cluster Heads | loss             |
|----+------------+-----------------+------------------|
|  0 |          0 |               7 | [1.035766720771] |
|  1 |          1 |               8 | [3.537582755088] |
|  2 |          2 |              14 | [2.12351319]     |
+----+------------+-----------------+------------------+

我的问题是如何将我的代码更新为 select 来自每个具有以下损失的集群（例如，3）的头。

结果看起来像

Tabel 2: heads of each  Cluster with loss below 3
+----+-----------+------------+---------+
|    |   element |   Clusters |    loss |
|----+-----------+------------+---------|
|  0 |         0 |          0 | 1.45316 |
|  2 |         2 |          2 | 2.22189 |
|  3 |         3 |          0 | 1.36913 |
|  4 |         4 |          0 | 1.29432 |
|  6 |         6 |          0 | 1.28239 |
|  7 |         7 |          0 | 1.03577 |
|  9 |         9 |          0 | 1.64741 |
| 10 |        10 |          2 | 2.12352 |
| 11 |        11 |          0 | 1.5225  |
| 12 |        12 |          0 | 1.09226 |
| 14 |        14 |          2 | 2.12351 |
+----+-----------+------------+---------+

当我尝试使用 min_losses = [[loss[elem]] for elem in elems if loss[elem] < 3] 我遇到以下错误：

---> 15 min_ind = min_losses.index(min_loss_temp)

ValueError: [[ 1.45316], [1.36913], [ 1.29432], [ 1.29432], [1.03577], [1.64741], [1.5225], [1.09226 ]] is not in list