Notes by Lex Toumbourou

A Convolutional Neural Network (CNN) is a particular type of neural network that learns feature engineer via filters (aka kernels) optimisation.

Typically a CNN contains multiple combinations of a [[../../../permanent/convolutional-layer]] and a [[Pooling Layer]].

Convolutional Layer

The convolutional layer extracts features from an input image by applying a [[Kernel]] to different parts of an image. The concept of a kernel comes from image processing , and is a small matrix which is applied over different regions of the matrix. It can be used for blurring, sharening, embossing, edge detection and so on.

Pooling Layer

The idea of a pooling layer is to pool different regions of the output image to reduce the dimensionality.