7

In the neural network community, is a tensor pretty much always just a multi-dimensional array?

amoeba
  • 93,463
  • 28
  • 275
  • 317
Franck Dernoncourt
  • 42,093
  • 30
  • 155
  • 271
  • At least for Theano and TensorFlow this seems to be the case. – Vladislavs Dovgalecs Nov 13 '15 at 06:47
  • Yes, in the same sense that a matrix is just a 2 dimensional array. But tensors just simplify notation and programming just as matrix operation simplifies multivariate methods. In particular for a convolutional net, you have 1 image (tile) being processed by a bank of filters, which can be expressed as a single tensor operation. – seanv507 Nov 13 '15 at 07:07

1 Answers1

3

Tensors in the neural network community = vector (1D-tensor), matrix/array (2D-tensor), or multi-dimensional array (nD-tensor, with $n > 2$).

Examples:

enter image description here

enter image description here


Related: Why the sudden fascination with tensors?

Franck Dernoncourt
  • 42,093
  • 30
  • 155
  • 271