WaveNet & SeriesNet - Avesta-Ahmadi/Univariate-Time-Series-Forecasting GitHub Wiki

WaveNet architecture is based on the residual networks and is widely used in speech recognition applications. In this project, this structure have been used for univariate time series forecasting. The following graph demonstrates the schematic of a WaveNet network.

Source: Oord, et al. WaveNet: A Generative Model for Raw Audio. ArXiv.org, 19 Sept. 2016, arxiv.org/abs/1609.03499.

Also, SeriesNet has been inspired by the WaveNet structure and modified to be used for univariate timeseries analysis. The structure of SeriesNet can be observed in the following figure.

Source: Borovykh, Anastasia, et al. Dilated Convolutional Neural Networks for Time Series Forecasting. Journal of Computational Finance, 2018, doi:10.21314/jcf.2019.358.