Multilayer Neural Networks

(5th-Nov-2020)

• Deeper architecture is more expressive than a shallow one – 1-layer nets only model linear hyper planes – 2-layer nets are universal function approximators: given infinite hidden nodes, it can express any continuous function.