Deep neural networks

A deep neural network is one that uses many hidden layers (2 or more) to answer complex questions. It does this by recursively breaking down the problem into smaller questions. The first layers answer simple questions, and each following layer uses that information to attempt to answer more abstract questions. They've been around since at least the '80's, but they are harder to train, so they've only become useful in the last 10 years or so as new methods have been developed to train them. Deep networks with 5-10 layers are now fairly common.

#ArtificialIntelligence #machinelearning #computerscience #programming

http://neuralnetworksanddeeplearning.com/chap1.html