DTW-NN: A novel neural network for time series recognition using dynamic alignment between inputs and weights

作者：

Highlights：

• Proposes Dynamic Time Warping Neural Network, a feed-forward network for time series.

• DTW-NN uses a temporal kernel-like function in replace of a typical inner product.

• DTW-NN uses dynamic programming to align weights and inputs.

• We evaluate on Unipen 1a, 1b, 1c, UCI Spoken Arabic, UCI ADL, and Flavia leaf shapes.

摘要

•Proposes Dynamic Time Warping Neural Network, a feed-forward network for time series.•DTW-NN uses a temporal kernel-like function in replace of a typical inner product.•DTW-NN uses dynamic programming to align weights and inputs.•We evaluate on Unipen 1a, 1b, 1c, UCI Spoken Arabic, UCI ADL, and Flavia leaf shapes.

论文关键词：Neural networks,Dynamic time warping,Temporal kernel,Time series,Dynamic programming

论文评审过程：Received 2 December 2018, Revised 14 August 2019, Accepted 19 August 2019, Available online 22 August 2019, Version of Record 20 January 2020.

论文官网地址：https://doi.org/10.1016/j.knosys.2019.104971