Article
Physics, Fluids & Plasmas
Tobias Braun, Cinthya N. Fernandez, Deniz Eroglu, Adam Hartland, Sebastian F. M. Breitenbach, Norbert Marwan
Summary: The analysis of irregularly sampled time series is a challenging task. We demonstrate that the edit distance is an effective metric for comparing time series segments of unequal length. We study the impact of sampling rate variations on recurrence quantification analysis and propose a method to correct for biases. The effectiveness of the proposed approach is demonstrated with an example and a real-world dataset.
Article
Computer Science, Artificial Intelligence
Zhen Wang, Ting Jiang, Zenghui Xu, Ji Zhang, Jianliang Gao
Summary: In this study, a multivariate temporal graph network is proposed to address three major challenges in modeling spatial and structural dependencies for IS-MTS data. The method utilizes a multivariate interaction module, a novel adjacent graph propagation mechanism, and a masked temporal-aware attention module to handle variable time intervals, asynchronous time points, and a lack of prior knowledge. Extensive experimental evaluation demonstrates the superior performance of the proposed method.
IEEE INTELLIGENT SYSTEMS
(2023)
Article
Biology
Lisa Steyer, Almond Stoecker, Sonja Greven
Summary: We present statistical analysis methods for samples of curves in two or more dimensions, focusing on the computation of means and distances. To address the challenges of irregular and sparse curve observations, we propose using spline curves and the elastic distance with a polygonal approach to ensure parameterization invariance. The effectiveness of our methods is demonstrated through two applications.
Article
Mathematics
Pablo Bonilla-Escribano, David Ramirez, Alejandro Porras-Segovia, Antonio Artes-Rodriguez
Summary: This study compares different variability metrics applied to irregularly sampled time series, identifying the most robust and interpretable measures. The results suggest a bias toward normalized and raw observation-based metrics in real data applications. Therefore, conclusions should be drawn using results from synthetic experiments.
Article
Mathematics, Applied
Qunxi Zhu, Xin Li, Wei Lin
Summary: This article proposes a data-driven and model-free approach for detecting unstable periodic orbits (UPOs) in chaotic systems based on irregularly sampled time series. The approach combines the neural differential equations (NDEs) and adaptive delayed feedback (ADF) technique, and can accurately reconstruct chaotic systems with irregular sampling rates, enhancing the detection ability of UPOs.
Article
Physics, Multidisciplinary
Celik Ozdes, Deniz Eroglu
Summary: Irregularly sampled time series analysis is a common problem in various disciplines. We propose a method to obtain a regularly sampled time series spectrum with minimum information loss. The approach is applied to validate a prototypical model and identify critical climate transition periods and their characteristic properties in palaeoclimate proxy data sets around Africa.
EUROPEAN PHYSICAL JOURNAL-SPECIAL TOPICS
(2023)
Article
Biochemical Research Methods
Sachin Heerah, Roberto Molinari, Stephane Guerrier, Amy Marshall-Colon
Summary: The study developed a statistical framework to identify causal gene-gene relationships in unevenly spaced, multivariate time series data. Utilizing maximum-likelihood estimation for parameter estimation, bootstrap procedures were used to test for Granger-Causality. The proposed approach identified 3078 significant interactions in Arabidopsis thaliana data, with 2012 interactions having root causal genes and 1066 interactions having shoot causal genes.
Article
Computer Science, Information Systems
Sindhu Tipirneni, Chandan K. Reddy
Summary: In this study, the authors propose a Self-supervised Transformer for Time-Series (STraTS) model to handle the challenges of sparsity and irregular time intervals in multivariate time-series data observed in critical care settings. The STraTS model treats time-series as a set of observation triplets and leverages a novel Continuous Value Embedding technique to encode continuous time and variable values. The model outperforms existing methods for mortality prediction, especially when labeled data is limited. Additionally, the authors present an interpretable version of STraTS that can identify important measurements in the time-series data.
ACM TRANSACTIONS ON KNOWLEDGE DISCOVERY FROM DATA
(2022)
Article
Multidisciplinary Sciences
Aditi Kathpalia, Pouya Manshour, Milan Palus
Summary: Distinguishing cause from effect is a scientific challenge, and Compression-Complexity Causality (CCC) and Permutation CCC (PCCC) propose new methods to address this challenge. CCC estimates causality based on changes in compression-complexity, and PCCC encodes multidimensional patterns into one-dimensional symbolic sequences. Both methods show promising results in dealing with irregular and uncertain data.
SCIENTIFIC REPORTS
(2022)
Article
Astronomy & Astrophysics
Lars T. Kreutzer, Edward Gillen, Joshua T. Briegal, Didier Queloz
Summary: We propose a generalized estimator, S-ACF, for the autocorrelation function that can effectively extract periodicity and signal shape information from time series. It works well for irregularly sampled data and requires minimal assumptions about the underlying process. The performance of S-ACF is comparable to commonly used estimators and has been demonstrated through various synthetic and astrophysical data sets.
MONTHLY NOTICES OF THE ROYAL ASTRONOMICAL SOCIETY
(2023)
Article
Environmental Sciences
Granville Tunnicliffe Wilson, John Haywood, Lynda Petherick
Summary: This study demonstrates how an autoregressive Gaussian process model incorporating a time scale coefficient can be used to represent irregularly sampled geophysical time series, providing flexibility appropriate to the data structure. The model improves the identification of periodicities and dependencies within and between series, leading to valuable insights into climate change predictions.
Article
Computer Science, Theory & Methods
Alexandre Constantin, Mathieu Fauvel, Stephane Girard
Summary: This paper investigates the classification of irregularly sampled Satellite image time-series (SITS) and proposes a multivariate Gaussian process mixture model to address the irregular sampling, multivariate nature, and scalability to large data-sets. The model handles spectral and temporal correlation using a Kronecker structure on the covariance operator of the Gaussian process. Experimental results on simulated and real SITS data highlight the significance of considering spectral correlation for achieving good classification accuracy and reconstruction errors.
STATISTICS AND COMPUTING
(2022)
Article
Mathematics, Applied
Jonghyeon Lee, Edward De Brouwer, Boumediene Hamzi, Houman Owhadi
Summary: A simple and interpretable method for learning a dynamical system from data is to interpolate its vector field using a kernel. This strategy, particularly efficient when using data-adapted kernels like Kernel Flows, breaks down when the observed time series is irregularly sampled. This paper proposes a solution to this problem by incorporating time differences between observations into the data-adapted kernels, resulting in an improved forecasting accuracy while maintaining simplicity, speed, and robustness.
PHYSICA D-NONLINEAR PHENOMENA
(2023)
Article
Geochemistry & Geophysics
Alexandre Constantin, Mathieu Fauvel, Stephane Girard
Summary: Recent satellite missions have generated a large amount of Earth observation data, with optical time series like Sentinel-2 or Landsat being used for land use and land cover studies. The proposed approach can deal with irregular temporal sampling and missing data, performing classification and reconstruction simultaneously. Experimental results show that the method yields robust reconstructions in the presence of undetected clouds or shadows, without requiring any temporal preprocessing.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2022)
Article
Physics, Fluids & Plasmas
William Davis, Bruce Buffett
Summary: This article presents extensions of two methods for estimating drift and diffusion functions from irregularly sampled time-series data. These methods are applicable to various stochastic systems and can be used for analyzing paleoclimatological isotope records.
Article
Mathematics, Interdisciplinary Applications
Bo Li, Tian Huang
Summary: This paper proposes an approximate optimal strategy based on a piecewise parameterization and optimization (PPAO) method for solving optimization problems in stochastic control systems. The method obtains a piecewise parameter control by solving first-order differential equations, which simplifies the control form and ensures a small model error.
CHAOS SOLITONS & FRACTALS
(2024)
Article
Mathematics, Interdisciplinary Applications
Guram Mikaberidze, Sayantan Nag Chowdhury, Alan Hastings, Raissa M. D'Souza
Summary: This study explores the collective behavior of interacting entities, focusing on the co-evolution of diverse mobile agents in a heterogeneous environment network. Increasing agent density, introducing heterogeneity, and designing the network structure intelligently can promote agent cohesion.
CHAOS SOLITONS & FRACTALS
(2024)
Article
Mathematics, Interdisciplinary Applications
Gengxiang Wang, Yang Liu, Caishan Liu
Summary: This investigation studies the impact behavior of a contact body in a fluidic environment. A dissipated coefficient is introduced to describe the energy dissipation caused by hydrodynamic forces. A new fluid damping factor is derived to depict the coupling between liquid and solid, as well as the coupling between solid and solid. A new coefficient of restitution (CoR) is proposed to determine the actual physical impact. A new contact force model with a fluid damping factor tailored for immersed collision events is proposed.
CHAOS SOLITONS & FRACTALS
(2024)