0

Suppose I have a neural network with fixed architecture. If I input one data point, the running time is 1s. Then what is the running time if the input is N data points?

Is it O(1) or O(N)? In other words, does the running time depend on the number of inputs in a linear way or constant? I heard the term "vectorization", which can reduce running time and is used in the feedforward neural network. But is vectorization constant in size?

Tony B
  • 272
  • 2
  • 8

1 Answers1

2

The running time is O(N), because you will have to do the same operations for each input.

Vectorization does not reduce the asymptotic complexity of the algorithm, it is just an implementation detail.

BlackBear
  • 22,411
  • 10
  • 48
  • 86