Machine Learning for Developers

Discrete convolution

Even when useful, the realm in which we work is eminently digital, so we need to translate this operation to the discrete domain. The convolution operation for two discrete functions f and g is the translation of the original integral to an equivalent summation in the following form:

This original definition can be applied to functions of an arbitrary number of dimensions. In particular, we will work on 2D images, because this is the realm of a large number of applications, which we will describe further in this chapter.

Now it's time to learn the way in which we normally apply the convolution operator, which is through ...