在Tensorflow中使用GMM时的Cholesky分解错误

Question

I face this strange problem while using GMM in Tensorflow.

InvalidArgumentError (see above for traceback): Cholesky decomposition was not successful. The input might not be valid.

The code I am using shows below.

from tensorflow.contrib.factorization.python.ops import gmm as gmm_lib
import random
import numpy as np
import tensorflow as tf

class DataGen:
    start = 0
    batch_size = 100
    end = start + batch_size

    def __init__(self, points):
        self.points = points

    def next_batch(self):
        start = self.start
        end = self.end
        if end <= self.points.shape[0]:
            self.start += self.batch_size
            self.end += self.batch_size
            return tf.constant(self.points[start: end]), None


mu, sigma = 0, 0.1
x_1d = 10*np.random.randn(500, 1).astype('f') + 50
data_gen = DataGen(x_1d)

gmm = gmm_lib.GMM(1,random_seed=0)
gmm.fit(input_fn=data_gen.next_batch)

Any idea to solving this problem? Thx.

Answer 1

Even I am facing the same trouble but that is not the case when we use google colaboratory !! You can perform all your experiment in the cloud within an environment alreay setup : [Colaboratory][1]

https://colab.research.google.com/notebooks/welcome.ipynb#scrollTo=-Rh3-Vt9Nev9