3

In machine learning concatenation seems to have $2$ different meanings depending on the context.

For 2 tensors $[a, b]$ and $[c,d]$, concatenations of these $2$ tensors seems to mean either

$[a,b,c,d]$ or $[a + c, b + d]$

Which is the correct interpretation of the concatenation operation?

Siong Thye Goh
  • 6,431
  • 3
  • 17
  • 28
Henry Chinner
  • 217
  • 2
  • 7
  • 1
    The Wikipedia lemma on [Concatenation](https://en.wikipedia.org/wiki/Concatenation) is quite straight-forward refer to that. (I feel for you; when authors overload existing terminology things can quickly become misleading) – usεr11852 May 28 '17 at 22:15

1 Answers1

3

The first one is correct, concatenation of tensors works exactly like string concatenation, except that on 2d or higher tensors you need to specify the dimension along which you're concatenating. Your second example is simply tensor addition.

rcpinto
  • 1,616
  • 12
  • 14