Differential radial basis function network for sequence modelling
作者：
Highlights：
• We propose a novel radial basis function network architecture termed RBF-DiffNet.
• RBF-DiffNet is fast and designed to improve generalisation in sequential data.
• The hidden layer blocks of RBF-DiffNet are partial differential equations (PDEs).
• The PDEs help regularise the network following modified backward-Euler updates.
• RBF-DiffNet outperforms LSTM and MLP ensembles on select artificial and real data.
摘要
•We propose a novel radial basis function network architecture termed RBF-DiffNet.•RBF-DiffNet is fast and designed to improve generalisation in sequential data.•The hidden layer blocks of RBF-DiffNet are partial differential equations (PDEs).•The PDEs help regularise the network following modified backward-Euler updates.•RBF-DiffNet outperforms LSTM and MLP ensembles on select artificial and real data.
论文关键词：Radial basis function,Neural network,Sequence modelling
论文评审过程：Received 20 January 2021, Revised 24 September 2021, Accepted 25 September 2021, Available online 13 October 2021, Version of Record 29 October 2021.
论文官网地址：https://doi.org/10.1016/j.eswa.2021.115982
