1

Given a convolution kernel, is there a standard way (hopefully some built-in function of some library) to get the (sparse) matrix that represents the convolution operation?

Specifically, I have a convolutional Keras layer from (54, 54, 128) to (24, 24, 128) and I want to get the matrix that represents the operation of this layer. Is there a standard way of doing this?

Nadav Schweiger
  • 97
  • 12
  • You should edit your question to make it absolutely clear about what you need, because the words you use make it easy to confuse with other terms. – Dr. Snoopy Jun 12 '19 at 14:26
  • @MatiasValdenegro Which word seems confusing? – user76284 Sep 25 '19 at 00:45
  • @user76284 Convolution matrix can be interpreted as the convolution kernel, the filters used to perform convolution on an image, but i think the OP wants the matrix that represents the operation from a linear algebra perspective. – Dr. Snoopy Sep 25 '19 at 00:51

0 Answers0