Deep Learning with Azure by Mathew Salvaris Danielle Dean & Wee Hyong Tok

Author:Mathew Salvaris, Danielle Dean & Wee Hyong Tok
Language: eng
Format: epub
Publisher: Apress, Berkeley, CA

The Convolution in Convolution Neural Networks

To keep things simple when talking about convolution we will be talking about discrete convolution. Mathematically, convolution is the simple summation of the pointwise multiplication of two functions. The summations can take place in one or more dimensions, so for grayscale images the summation would take place over two dimensions and over three dimensions in color images.

Convolution is similar to cross-correlation and in many deep learning libraries the implementation is actually cross-correlation even though it is referred to as convolution. For all practical purposes in the CNNs this is just an implementation detail and does not really affect the resulting behavior of the model. To get an intuitive feeling of how convolution behaves, there is a simple example illustrated in Figure 6-1.

Figure 6-1Convolution in CNNs

Download

Copyright Disclaimer:
This site does not store any files on its server. We only index and link to content provided by other sites. Please contact the content providers to delete copyright contents if any and email us, we'll remove relevant links or contents immediately.