A multi-layer neural network is a nueral betwork with more than two layers (input and output). The earliest form was the multi-layer perceptron with three layers (input-hidden-output) while deep neural networks have many layers. .While the development of backpropagation meant that deeper neural networks were theoretically possible from the alet 1980s, it was only the availability of large data and massive computation in the mid 2010s, that made deep neural networks practical.
Used in Chap. 6: pages 89, 92; Chap. 9: pages 131, 132
A multi-layer perceptron architecture.
Deep learning architecture – multiple layers, with varying connection topologies.