Dynamic Recurrent Neural Networks
DRNN are a class of fully recurrent neural networks
endowed with some internal memory.
By virtue of their internal dynamics, they approximate the underlying law governing a time series by a system of
non-linear equations of internal variables.