如何计算意味着在Tensorflow中通过indice进行分组?

时间:2017-10-02 18:35:40

标签: python numpy tensorflow aggregate k-means

为了实现k-means算法,我为每个样本计算了形状using System; namespace caesarCipher { class Program { static void Main(string[] args) { string text; Console.WriteLine("Enter the text to encrypt "); text = System.Convert.ToString(Console.ReadLine()); string lower = text.ToLower(); Random rnd = new Random(); int shift = rnd.Next(1, 25); foreach (char c in lower) { int unicode = c; int shiftUnicode = unicode + shift; Console.WriteLine(shiftUnicode); if (shiftUnicode >= 123) { int overflowUnicode = 97 + (shiftUnicode - 123); char character = (char)overflowUnicode; string newText = character.ToString(); } else { char character = (char)shiftUnicode; string newText = character.ToString(); } } Console.ReadLine(); } } } 的聚类索引0..K-1

(N, 1)

现在我想采用样本张量0 1 2 2 1 1 ... 0 并计算新的形状(N, 3)

(K, 3)我会写:

numpy

如何对for i_cluster in range(n_clusters): mu[i_cluster, :] = X[cluster_indice == i_cluster, :].mean(0) 执行相同操作?

更新

我可能需要使用tf.dynamic_partition() ...

1 个答案:

答案 0 :(得分:0)

我这样写道:

parent::__construct($par1)