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

Model 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 with relu activation function.
  • A SkipGRU similar to the previous GRU cell, with the difference that the hidden state is taken from a specific amount of timesteps back in time. Both the GRU and the SkipGRU 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.