Neural Network Programming with TensorFlow

TensorFlow comparison with Numpy

TensorFlow and Numpy are both N-dimensional array libraries. TensorFlow additionally allows us to create tensor functions and compute derivatives. TensorFlow has become one of the major libraries used for deep learning as it is incredibly efficient and can run on GPUs.

The following program describes how TensorFlow and numpy can be used to perform similar operations such as creating tensors of a (3,3) shape:

import TensorFlow  as tfimport numpy as nptf.InteractiveSession()# TensorFlow  operationsa = tf.zeros((3,3))b = tf.ones((3,3))print(tf.reduce_sum(b, reduction_indices=1).eval())print(a.get_shape())# numpy operationsa = np.zeros((3, 3))b = np.ones((3, 3))print(np.sum(b, axis=1))
print(a.shape)

The output ...