Published on Thu Feb 14 2019

Sinkhorn Divergence of Topological Signature Estimates for Time Series Classification

Colin Stephen

Distinguishing between classes of time series sampled from dynamic systems is a common challenge in systems and control engineering. In this paper we address these issues with a new non parametric classifier based on topological signatures. Our model learns classes as weighted kernel density estimates over persistent homology diagrams.

0
0
0
Abstract

Distinguishing between classes of time series sampled from dynamic systems is a common challenge in systems and control engineering, for example in the context of health monitoring, fault detection, and quality control. The challenge is increased when no underlying model of a system is known, measurement noise is present, and long signals need to be interpreted. In this paper we address these issues with a new non parametric classifier based on topological signatures. Our model learns classes as weighted kernel density estimates (KDEs) over persistent homology diagrams and predicts new trajectory labels using Sinkhorn divergences on the space of diagram KDEs to quantify proximity. We show that this approach accurately discriminates between states of chaotic systems that are close in parameter space, and its performance is robust to noise.

Wed Dec 11 2019
Machine Learning
The Wasserstein-Fourier Distance for Stationary Time Series
The Wasserstein-Fourier distance (WF) is used to measure the similarity between time series. The distance is calculated by calculating the distance between the (normalised) power spectral densities (NPSD) of time series and their normalised frequencies.
0
0
0
Sat Oct 10 2020
Machine Learning
TOTOPO: Classifying univariate and multivariate time series with Topological Data Analysis
0
0
0
Fri Feb 07 2020
Machine Learning
Unsupervised non-parametric change point detection in quasi-periodic signals
New unsupervised and non-parametric method to detect change in intricate quasi-periodic signals. Algorithm designed to detect changes in virtually any harmonic or a partially harmonic signal. Found abnormal or irregular cardiac cycles in waveforms for six of the most frequent clinical
0
0
0
Wed Feb 03 2021
Machine Learning
Time Series Classification via Topological Data Analysis
In this paper, we develop topological data analysis methods for classification tasks on univariate time series. We perform binary and ternary classification tasks on two public datasets that consist of physiological signals collected under stress and non-stress conditions.
0
0
0
Mon Mar 06 2017
Machine Learning
A time series distance measure for efficient clustering of input output signals by their underlying dynamics
This paper tackles the problem of automatically clustering these time series. We propose an extension to the so-called Martin cepstral distance. We show that the new distance measure performs as good as when every input/output pair is modeled explicitly.
0
0
0
Fri Dec 02 2016
Machine Learning
A General Framework for Density Based Time Series Clustering Exploiting a Novel Admissible Pruning Strategy
Time Series Clustering is an important subroutine in many higher-level data mining analyses. The superiority of Dynamic Time Warping over Euclidean distance gradually diminishes as we consider ever larger datasets. We propose a novel pruning strategy that exploits both the upper and lower bounds of DTW.
0
0
0