Depth of a network

The number of layers in the network.

Feature vector / representation / volume

A three dimensional tensor of size $W&space;\times&space;H&space;\times&space;D$ obtained in a certain layer of a neural network. W is the width, H is the height and D is the depth, i.e., the number of channels. If there is more than one example, this becomes a four dimensional tensor of size $W&space;\times&space;H&space;\times&space;D&space;\times&space;B$ , where $B$ is the batch size. image source

Spatial invariant feature vector

A feature vector that remains unchanged even if the input to the network is spatially translated.

Filters and biases

Filters are a four dimensional tensor of size $F&space;\times&space;F&space;\times&space;D&space;\times&space;K$ and biases are a vector of length $K$ . $F$ is the width and height of the filter, $D$ is the number of channels and $K$ is the number of filters.

Neighbourhood

A group of consecutive entries in a two-dimensional signal that has a rectangular or a square shape. image source

back