LSTnet
This "Long- and Short-term Time-series network" is based on the paper by Lai et. al.. See also the corresponding blog post.
Network Structure
Image from Lai et al, "Long- and Short-term Time-series network", ArXiv 2017.
The neural net consists of the following elements:
- A convolutional layer than operates on some window of the time series.
- A
GRU
cell withrelu
activation function. - A
SkipGRU
similar to the previousGRU
cell, with the difference that the hidden state is taken from a specific amount of timesteps back in time. Both theGRU
and theSkipGRU
layer take their input from the convolutional layer. - A dense layer that operates on the concatenated output of the previous two layers.
- An autoregressive layer operating on the input data itself, being added to the output of the dense layer.