Neural networks with many layers.
THE DEFINITION
Machine learning using layered networks loosely inspired by neurons.
WHY DEEP
Many layers. Early layers detect simple features; later ones combine them into complex ones.
WHAT MADE IT POSSIBLE
Very large amounts of data, substantially more computation, and better training methods.
Those three arriving together produced the advances of the last decade.
WHAT IT ENABLED
Image recognition at and beyond human level Usable speech recognition Machine translation that reads naturally Language models
WHAT DISTINGUISHES IT
The system learns which features matter rather than being told.
Nobody defines what an edge is; the system discovers edges are useful.
WHAT IT DID NOT SOLVE
Reliable reasoning Knowing what it does not know Learning from few examples
RELATED TERMS
Neural network, transformer, parameter.
