Convolutional Neural Networks

Question 1

Why do we want to use CNN?

Answer 1

Simple neural nets can become super complex when used on images, so we need a better way. Deep neural nets are an option, with CNN being the most famous.

Question 2

What are the building blocks of CNN?

Answer 2

Convolution - move kernel over pixels
Activation - takes output from convolution and passes
through functions. This speeds up the training and
introductes non-linearities
Pooling - Downsampling of features, keep the most
important data