# Tensor flow error - conversion and peformance

Here is my code:

import tensorflow as tf
import numpy as np
import matplotlib.pyplot as plt

x = np.linspace(0,10,20)
train_x = np.array([[i] for i in x])
train_y = np.sin(x)

regressor = tf.contrib.learn.DNNRegressor(hidden_units=[10,20, 10])
regressor.fit(x = train_x, y = train_y, steps = 2000)
predictions = regressor.predict(x = train_x)

plt.plot(x, train_y)
plt.plot(x, predictions)
plt.show()


It generates the following error message/warning:

/usr/bin/python3.4 /home/ttt/Dropbox/Programming/Python/Tensor_flow/one_dim_case.py
/usr/local/lib/python3.4/dist-packages/tensorflow/python/ops/array_ops.py:1197:
VisibleDeprecationWarning: converting an array with ndim > 0 to an
index will result in an error in the future
result_shape.insert(dim, 1)

Process finished with exit code 0


Could you help me to understand what is wrong?

An additional remark, with so many hidden layers and neurons and number of iterations, I really surprised with poor sin approximation.

P.S. This is a toy example to help to understand tensorflow