Sequence Prediction using Spectral RNNs

Moritz Wolter, Juergen Gall und Angela Yao
In proceedings of International Conference on Artificial Neural Networks, Springer, 2020
 

Abstract

Fourier methods have a long and proven track record as an excellent tool in data processing. As memory and computational constraints gain importance in embedded and mobile applications, we propose to combine Fourier methods and recurrent neural network architectures. The short-time Fourier transform allows us to efficiently process multiple samples at a time. Additionally, weight reduction trough low pass filtering is possible. We predict time series data drawn from the chaotic Mackey-Glass differential equation, real-world power load and motion capture data.

Stichwörter: Frequency Domain, Sequence Modelling, Short Time Fourier Transform

Source code available online at https://github.com/v0lta/Spectral-RNN .

Bilder

Bibtex

@INPROCEEDINGS{wolter2020spectral,
     author = {Wolter, Moritz and Gall, Juergen and Yao, Angela},
      title = {Sequence Prediction using Spectral RNNs},
  booktitle = {International Conference on Artificial Neural Networks},
       year = {2020},
  publisher = {Springer},
   keywords = {Frequency Domain, Sequence Modelling, Short Time Fourier Transform},
   abstract = {Fourier methods have a long and proven track record as an excellent tool in data processing.  As
               memory and computational constraints gain importance in embedded and mobile applications, we propose
               to combine Fourier methods and recurrent neural network architectures.   The short-time Fourier
               transform allows us to efficiently process multiple samples at a time.  Additionally, weight
               reduction trough low pass filtering is possible. We predict time series data drawn from the chaotic
               Mackey-Glass differential equation, real-world power load and motion capture data.},
        url = {https://arxiv.org/pdf/1812.05645.pdf}
}