4.6 Article

Transition Manifolds of Complex Metastable Systems

期刊

JOURNAL OF NONLINEAR SCIENCE
卷 28, 期 2, 页码 471-512

出版社

SPRINGER
DOI: 10.1007/s00332-017-9415-0

关键词

Metastability; Reaction coordinate; Coarse graining; Effective dynamics; Whitney embedding theorem; Transfer operator

资金

  1. Deutsche Forschungsgemeinschaft (DFG) [CRC 1114]
  2. Einstein Foundation Berlin (Einstein Center ECMath)

向作者/读者索取更多资源

We consider complex dynamical systems showing metastable behavior, but no local separation of fast and slow time scales. The article raises the question of whether such systems exhibit a low-dimensional manifold supporting its effective dynamics. For answering this question, we aim at finding nonlinear coordinates, called reaction coordinates, such that the projection of the dynamics onto these coordinates preserves the dominant time scales of the dynamics. We show that, based on a specific reducibility property, the existence of good low-dimensional reaction coordinates preserving the dominant time scales is guaranteed. Based on this theoretical framework, we develop and test a novel numerical approach for computing good reaction coordinates. The proposed algorithmic approach is fully local and thus not prone to the curse of dimension with respect to the state space of the dynamics. Hence, it is a promising method for data-based model reduction of complex dynamical systems such as molecular dynamics.

作者

我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。

评论

主要评分

4.6
评分不足

次要评分

新颖性
-
重要性
-
科学严谨性
-
评价这篇论文

推荐

Article Mathematics, Applied

Data-driven approximation of the Koopman generator: Model reduction, system identification, and control

Stefan Klus, Feliks Nuske, Sebastian Peitz, Jan-Hendrik Niemann, Cecilia Clementi, Christof Schuette

PHYSICA D-NONLINEAR PHENOMENA (2020)

Article Physics, Multidisciplinary

Kernel-Based Approximation of the Koopman Generator and Schrodinger Operator

Stefan Klus, Feliks Nueske, Boumediene Hamzi

ENTROPY (2020)

Article Physics, Multidisciplinary

Spectral Properties of Effective Dynamics from Conditional Expectations

Feliks Nueske, Peter Koltai, Lorenzo Boninsegna, Cecilia Clementi

Summary: This paper investigates the reduction of high-dimensional systems to effective models through the conditioning approach, comparing the spectrum of the generator of the resulting effective dynamics to the full generator. A new relative error bound for reversible systems is proven in terms of eigenfunction approximation error. Numerical examples suggest that Kramers-Moyal type approximations for computing the spectrum of the reduced generator are largely insensitive to the time window used for the estimators, with implications for underdamped Langevin dynamics.

ENTROPY (2021)

Article Mathematics, Applied

Dimensionality Reduction of Complex Metastable Systems via Kernel Embeddings of Transition Manifolds

Andreas Bittracher, Stefan Klus, Boumediene Hamzi, Peter Koltai, Christof Schuette

Summary: In this article, a novel kernel-based machine learning algorithm is presented for identifying the low-dimensional geometry of the effective dynamics of high-dimensional multiscale stochastic systems. By embedding and learning the transition manifold in a reproducing kernel Hilbert space, the approach is enhanced and shown to preserve the manifold structure under the embedding with distortion bounds. This results in a more robust and efficient algorithm compared to previous parameterization approaches.

JOURNAL OF NONLINEAR SCIENCE (2021)

Article Meteorology & Atmospheric Sciences

Estimation of Koopman Transfer Operators for the Equatorial Pacific SST

Antonio Navarra, Joe Tribbia, Stefan Klus

Summary: Ensemble methods have become popular tools in atmospheric, climate, and ocean dynamics studies for obtaining statistical information on systems and approximating probability distributions. This paper introduces a method to directly estimate probability distribution evolution and create predictor systems based on nonlinear formulations, showcasing its application in two examples. The objective is to demonstrate the utility of these methods in complex, multidimensional atmosphere and ocean environments.

JOURNAL OF THE ATMOSPHERIC SCIENCES (2021)

Article Multidisciplinary Sciences

Data-driven model reduction of agent-based systems using the Koopman generator

Jan-Hendrik Niemann, Stefan Klus, Christof Schuette

Summary: This paper demonstrates how Koopman operator theory can be used to derive reduced models of agent-based systems using only simulation data. The goal is to learn coarse-grained models and represent the reduced dynamics using ordinary or stochastic differential equations. Experimental results show that the obtained reduced systems are in good agreement with analytical results when the number of agents is sufficiently large.

PLOS ONE (2021)

Article Computer Science, Artificial Intelligence

Feature space approximation for kernel-based supervised learning

Patrick Gelss, Stefan Klus, Ingmar Schuster, Christof Schuette

Summary: The proposed method aims to approximate high- or even infinite-dimensional feature vectors to reduce training data size, storage consumption, and computational complexity, while enhancing the generalizability of learned target functions. Significant improvements have been demonstrated in classification and regression problems across various application areas.

KNOWLEDGE-BASED SYSTEMS (2021)

Article Mathematics, Applied

Tensor-based computation of metastable and coherent sets

Feliks Nueske, Patrick Gelss, Stefan Klus, Cecilia Clementi

Summary: Recent years have witnessed significant progress in data-driven analysis of dynamical systems based on Koopman operator theory and the TT format for solving large-scale problems. Combining these two approaches enables handling high-dimensional problems with rich basis functions or feature sets, and deriving efficient algorithms for reducing the system's evolution operator representation.

PHYSICA D-NONLINEAR PHENOMENA (2021)

Article Physics, Multidisciplinary

Koopman analysis of quantum systems*

Stefan Klus, Feliks Nueske, Sebastian Peitz

Summary: Koopman operator theory has wide applications in various research areas and this paper demonstrates the use of data-driven methods to analyze quantum physics problems and solve the Schrödinger equation, opening up a new avenue for research.

JOURNAL OF PHYSICS A-MATHEMATICAL AND THEORETICAL (2022)

Article Mathematics, Applied

Koopman-Based Spectral Clustering of Directed and Time-Evolving Graphs

Stefan Klus, Natasa Djurdjevac Conrad

Summary: This paper discusses the challenges of clustering directed graphs and proposes a clustering algorithm based on the concept of metastable sets.

JOURNAL OF NONLINEAR SCIENCE (2023)

Article Mathematics

Overcoming the timescale barrier in molecular dynamics: Transfer operators, variational principles and machine learning

Christof Schuette, Stefan Klus, Carsten Hartmann

Summary: This article discusses how to overcome the 'timescale barrier' in molecular dynamics through transfer operator-based techniques. It focuses on the approximation of dynamical behavior on long timescales, the theory and algorithmic development of transfer operators, and their relation to rare event simulation techniques. The article takes a mathematical perspective and refers to existing research articles for applications in real-world molecular systems.

ACTA NUMERICA (2023)

Article Chemistry, Physical

Efficient approximation of molecular kinetics using random Fourier features

Feliks Nueske, Stefan Klus

Summary: The use of random Fourier features as a stochastic approximation method allows for more efficient estimation of slow kinetic processes.

JOURNAL OF CHEMICAL PHYSICS (2023)

Article Computer Science, Artificial Intelligence

Deeptime: a Python library for machine learning dynamical models from time series data

Moritz Hoffmann, Martin Scherer, Tim Hempel, Andreas Mardt, Brian de Silva, Brooke E. Husic, Stefan Klus, Hao Wu, Nathan Kutz, Steven L. Brunton, Frank Noe

Summary: Deeptime is a Python library that offers various tools for estimating dynamical models based on time-series data and provides analysis methods to compute important properties of the system. It is easy to use and extend, and compatible with scikit-learn.

MACHINE LEARNING-SCIENCE AND TECHNOLOGY (2022)

Article Computer Science, Artificial Intelligence

Symmetric and antisymmetric kernels for machine learning problems in quantum physics and chemistry

Stefan Klus, Patrick Gelss, Feliks Nueske, Frank Noe

Summary: This paper explores symmetric and antisymmetric kernels by analyzing their properties and computing the feature space dimensions of resulting polynomial kernels. It proves that the reproducing kernel Hilbert spaces induced by symmetric and antisymmetric Gaussian kernels are dense in the space of symmetric and antisymmetric functions. Additionally, a Slater determinant representation of the antisymmetric Gaussian kernel is proposed for efficient evaluation in high-dimensional state spaces, and shows that the size of the training data set can be significantly reduced by exploiting symmetries or antisymmetries.

MACHINE LEARNING-SCIENCE AND TECHNOLOGY (2021)

Article Computer Science, Theory & Methods

GraphKKE: graph Kernel Koopman embedding for human microbiome analysis

Kateryna Melnyk, Stefan Klus, Gregoire Montavon, Tim O. F. Conrad

APPLIED NETWORK SCIENCE (2020)

暂无数据