Nonsmooth Analysis and Subgradient Methods for Averaging in Dynamic Time Warping Spaces

01/23/2017
by   David Schultz, et al.
0

Time series averaging in dynamic time warping (DTW) spaces has been successfully applied to improve pattern recognition systems. This article proposes and analyzes subgradient methods for the problem of finding a sample mean in DTW spaces. The class of subgradient methods generalizes existing sample mean algorithms such as DTW Barycenter Averaging (DBA). We show that DBA is a majorize-minimize algorithm that converges to necessary conditions of optimality after finitely many iterations. Empirical results show that for increasing sample sizes the proposed stochastic subgradient (SSG) algorithm is more stable and finds better solutions in shorter time than the DBA algorithm on average. Therefore, SSG is useful in online settings and for non-small sample sizes. The theoretical and empirical results open new paths for devising sample mean algorithms: nonsmooth optimization methods and modified variants of pairwise averaging methods.

READ FULL TEXT

page 19

page 20

research
10/14/2016

A Reduction Theorem for the Sample Mean in Dynamic Time Warping Spaces

Though the concept of sample mean in dynamic time warping (DTW) spaces i...
research
09/30/2019

An Average-Compress Algorithm for the Sample Mean Problem under Dynamic Time Warping

Computing a sample mean of time series under dynamic time warping (DTW) ...
research
09/07/2018

Revisiting Inaccuracies of Time Series Averaging under Dynamic Time Warping

This article revisits an analysis on inaccuracies of time series averagi...
research
09/02/2021

Accurate shape and phase averaging of time series through Dynamic Time Warping

We propose a novel time series averaging method based on Dynamic Time Wa...
research
06/22/2020

Aligning Time Series on Incomparable Spaces

Dynamic time warping (DTW) is a useful method for aligning, comparing an...
research
11/03/2015

Properties of the Sample Mean in Graph Spaces and the Majorize-Minimize-Mean Algorithm

One of the most fundamental concepts in statistics is the concept of sam...
research
12/10/2019

Adaptive Dynamic Model Averaging with an Application to House Price Forecasting

Dynamic model averaging (DMA) combines the forecasts of a large number o...

Please sign up or login with your details

Forgot password? Click here to reset