TensorFlowでk-meansを実装

現在 Learning TensorFlow :: Clustering and k-meansで学習しているのですが，このアルゴリズムで10回ないし特定の閾値を超えるまでk-meansを繰り返したいとき，どのようにすればいいのでしょうか？

上述のサイトと同様のコードですが，以下にコードを掲出します。

functions.py

python
1import tensorflow as tf
2import numpy as np
3
4def create_samples(n_clusters, n_samples_per_cluster, n_features, embiggen_factor, seed):
5    np.random.seed(seed)
6    slices = []
7    centroids = []
8    for i in range(n_clusters):
9        samples = tf.random_normal((n_samples_per_cluster, n_features),
10                                   mean=0.0, stddev=5.0, dtype=tf.float32, seed=seed, name="cluster_{}".format(i))
11        current_centroid = (np.random.random((1, n_features)) * embiggen_factor) - (embiggen_factor/2)
12        centroids.append(current_centroid)
13        samples += current_centroid
14        slices.append(samples)
15    samples = tf.concat(slices, 0, name='samples')
16    centroids = tf.concat(centroids, 0, name='centroids')
17    return centroids, samples
18
19def plot_clusters(all_samples, centroids, n_samples_per_cluster):
20    import matplotlib.pyplot as plt
21    colour = plt.cm.rainbow(np.linspace(0,1,len(centroids)))
22    for i, centroid in enumerate(centroids):
23         samples = all_samples[i*n_samples_per_cluster:(i+1)*n_samples_per_cluster]
24         plt.scatter(samples[:,0], samples[:,1], c=colour[i])
25         plt.plot(centroid[0], centroid[1], markersize=35, marker="x", color='k', mew=10)
26         plt.plot(centroid[0], centroid[1], markersize=30, marker="x", color='m', mew=5)
27    plt.show()
28
29def choose_random_centroids(samples, n_clusters, seed=None):
30    n_samples = tf.shape(samples)[0]
31    random_indices = tf.random_shuffle(tf.range(0, n_samples), seed=seed)
32    begin = [0,]
33    size = [n_clusters,]
34    size[0] = n_clusters
35    centroid_indices = tf.slice(random_indices, begin, size)
36    initial_centroids = tf.gather(samples, centroid_indices)
37    return initial_centroids
38
39def assign_to_nearest(samples, centroids):
40    expanded_vectors = tf.expand_dims(samples, 0)
41    expanded_centroids = tf.expand_dims(centroids, 1)
42    distances = tf.reduce_sum( tf.square(
43        tf.subtract(expanded_vectors, expanded_centroids)), 2)
44    mins = tf.argmin(distances, 0)
45
46    nearest_indices = mins
47    return nearest_indices
48
49def update_centroids(samples, nearest_indices, n_clusters):
50    nearest_indices = tf.to_int32(nearest_indices)
51    partitions = tf.dynamic_partition(samples, nearest_indices, n_clusters)
52    new_centroids = tf.concat([tf.expand_dims(tf.reduce_mean(partition, 0), 0) for partition in partitions], 0)
53    return new_centroids

generate_samples.py

import tensorflow as tf
import numpy as np

from functions import *

n_features = 2
n_clusters = 5
n_samples_per_cluster = 500
seed = 700
embiggen_factor = 70

data_centroids, samples = create_samples(n_clusters, n_samples_per_cluster, n_features, embiggen_factor, seed)
initial_centroids = choose_random_centroids(samples, n_clusters)
nearest_indices = assign_to_nearest(samples, initial_centroids)
updated_centroids = update_centroids(samples, nearest_indices, n_clusters)


model = tf.global_variables_initializer()
with tf.Session() as session:
    sample_values = session.run(samples)
    updated_centroid_value = session.run(updated_centroids)
    print(updated_centroid_value)

    plot_clusters(sample_values, updated_centroid_value, n_samples_per_cluster)

nearest_indices および updated_centroids を繰り返し実行するようにすればいいのだと思うのですが，どのようにして更新されたセントロイドの情報を保持し，それをもとに nearest_indices を実行すればよいのか悩んでいます。

よろしくお願いします。