BP is a forwardbroadcasting multiply-layer-network that trains the weight of the differentiable nonlinear functions.

• BP是对非线性可微分函数进行权值训练的多层前向网络。