Summary of artificial neural network in a word
Summary of artificial neural network in a word
Core: a multi-layer composite function.
Artificial neural network is essentially a multi-layer compound function:
It realizes the mapping from vector x to vector y. Due to the use of the nonlinear activation function f, this function is a nonlinear function.
The problem solved in neural network training is not a convex optimization problem. The back propagation algorithm is derived from the chain rule of derivation of multivariate compound functions.
The standard neural network is a supervised learning algorithm and a nonlinear model, which can be used not only for classification problems, but also for regression problems.