Nonsmooth analysis and subgradient methods for averaging in dynamic time warping spaces

作者：

Highlights：

• Stochastic subgradient method outperforms state-of-the-art algorithm DTW Barycenter Averaging (DBA) for non-small sample sizes.

• Necessary and sufficient conditions of optimality for the sample mean problem in DTW spaces.

• Finite convergence of DBA to necessary conditions of optimality.

摘要

•Stochastic subgradient method outperforms state-of-the-art algorithm DTW Barycenter Averaging (DBA) for non-small sample sizes.•Necessary and sufficient conditions of optimality for the sample mean problem in DTW spaces.•Finite convergence of DBA to necessary conditions of optimality.

论文关键词：Dynamic time warping,Time series averaging,Sample mean,Fréchet function,Subgradient methods

论文评审过程：Received 20 January 2017, Revised 5 July 2017, Accepted 9 August 2017, Available online 20 September 2017, Version of Record 2 October 2017.

论文官网地址：https://doi.org/10.1016/j.patcog.2017.08.012