Convolutions
Get an intuitive understanding of convolutions using Jeremies Excel-Tools. We see how the matrix is constructed (basically repeated matrix multiplications). We also see how to optimize the filters.
Convolutions, explained by 3Blue1Brown: → Video
Pooling
The pooling operation involves sliding a two-dimensional filter over each channel of feature map and summarizing the features lying within the region covered by the filter → normally not used anymore!
→ Explains when to use max polling (when the thing we want to detect in the picture is tiny) and when to use average polling.
Convolutions as matrix products
As said above, convolutions are basically many different matrix multiplications. We look at it in more detail and see we could do the same with a single matrix multiplication!
Dropout
A simple way to prevent Neural Networks from Overfitting . We remove parts of the image based on the dropout filter.
Dropout regularization is a technique to prevent neural networks from overfitting. Dropout works by randomly disabling neurons and their corresponding connections. This prevents the network from relying too much on single neurons and forces all neurons to learn to generalize better.