A Multi-strategy LSHADE Algorithm and its Applications on Temporal Alignment

EasyChair Preprint no. 3480

Abstract

Temporal alignment is a common problem, but performance of corresponding techniques is easily affected by time series characteristic. In this paper, we propose a novel temporal alignment approach with a combination of multi-strategy success-history based adaptive differential evolution with linear population size reduction (MLSHADE) and dynamic time warping (DTW), namely, MLDTW. To strengthen efficiency of optimization, weighted mutation strategy, inferior solution search strategy and eigen Gaussian random walk strategy is presented. Moreover, the MLDTW operators which include initialization and updating operators are given to enhance the accuracy and speed of DTW based on MLSHADE. The performance of MLSHADE is verified by using CEC 2018 test suite. Meanwhile, the availability and robustness of MLDTW is proved by using sinusoidal signal and UCR time series datasets.

Keyphrases: CEC 2018, Dynamic Time Warping, Evolutionary Computation, multi-strategy, Temporal alignment

@Booklet{EasyChair:3480,
  author = {Zhenglei Wei and Changqiang Huang},
  title = {A Multi-strategy LSHADE Algorithm and its Applications on Temporal Alignment},
  howpublished = {EasyChair Preprint no. 3480},

  year = {EasyChair, 2020}}