Definition
A machine-learning model built from connected computational units, parameters and nonlinear transformations.
Usage and distinctions
Neural networks support tasks such as recognition, prediction and generation. The discussion contrasts writing or modifying parameters directly with optimizing them through training.