1
$\begingroup$

The output of a hidden layer $h_i$ in a convolutional neural network is (generally) a 3D grid of values. These values are the outputs of the neurons of layer $h_i$. Is there a standard way to refer to this matrix of values?

$\endgroup$
2
  • $\begingroup$ what you mean by: Is there a standard way to refer to this matrix of values? $\endgroup$ – Hunar Jun 8 '19 at 11:20
  • $\begingroup$ Is there some standard terminology for the matrix of values, something like 'activation matrix'. $\endgroup$ – sanso Jun 8 '19 at 11:33
0
$\begingroup$

It is called a feature map.

This image might be helpful for terminology:

convolution of

$\endgroup$
0
$\begingroup$

Not that I’m aware. The 2-d matrices indexed by the third dimension are often called channels. Other than that, “intermediate representations”, maybe? But since it isn’t really standard, you should explain whatever you decide to call it, to make sure it’s unambiguous.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.