neural-network architecture

Terms from Artificial Intelligence: humans at the heart of algorithms

Page numbers are for draft copy at present; they will be replaced with correct numbers when final book is formatted. Chapter numbers are correct and will not change now.

Deep neural networks may have many layers in different patterns of connection within and between kayers and different learning rules. For example often a convolutional neural network may form the first stage of an image processing network. The choice of the number of layers, their types and sizes comprises the neural network architecture. Typically this is designed by a human, but sometimes some of the hyperparameters, such as the size of a pinch-point layer, may be tuned by a meta-level algorithm.

Used in Chap. 6: page 92; Chap. 8: page 113; Chap. 12: page 197; Chap. 14: page 231

Also known as architecture

Deep learning architecture – multiple layers, with varying connection topologies.

Convolutional neural network – often used as the first stage of an image processing neural netwrok

Different kinds of connection patterns between layers.