\`x^2+y_1+z_12^34\`
Advanced Search
Article Contents
Article Contents

Nested smoothing algorithms for inference and tracking of heterogeneous multi-scale state-space systems

  • *Corresponding author: Sara Pérez-Vieites

    *Corresponding author: Sara Pérez-Vieites 
Abstract / Introduction Full Text(HTML) Figure(6) Related Papers Cited by
  • Multi-scale problems, where variables of interest evolve in different time-scales and live in different state-spaces, can be found in many fields of science. Here, we introduce a new recursive methodology for Bayesian inference that aims at estimating the static parameters and tracking the dynamic variables of these kind of systems. Although the proposed approach works in rather general setups, for clarity we analyze the case of a heterogeneous multi-scale model with 3 time-scales (static parameters, slow dynamic state variables and fast dynamic state variables). The proposed scheme, based on a nested filtering methodology of [27], combines three intertwined layers of filtering techniques that approximate recursively the joint posterior probability distribution of the parameters and both sets of dynamic state variables given a sequence of partial and noisy observations. We explore the use of both sequential Monte Carlo schemes and several Kalman filtering techniques in the different layers of the methodology to obtain approximations of the posterior probability distributions of interest. Some numerical results are presented for a stochastic two-scale Lorenz 96 model with unknown parameters.

    Mathematics Subject Classification: Primary: 62F15, 62M20; Secondary: 62M05.

    Citation:

    \begin{equation} \\ \end{equation}
  • 加载中
  • Figure 1.  Schematic depiction of the optimal smoother. Each column represents a layer of computation and the dependencies among pdfs are indicated by arrows. The dashed arrows are used to show relations among different layers while the solid arrows represent dependencies in the same layer.

    Figure 2.  Averaged NMSE$ _{ \boldsymbol{\theta}} $ (plot 2a), NMSE$ _{ \boldsymbol{x}} $ (plot 2b), NMSE$ _{ \boldsymbol{z}} $ (plot 2c) and average running time (plot 2d) of NPF with state augmentation (in green), SMC-SMC-UKF (in blue) and SMC-EnKF-EKF (in red), averaged over 50 simulation runs. The number of particles of the second layer of every method is set to $ J = N $..

    Figure 3.  Normalized $ {\widehat{\text{ESS}}_{ \boldsymbol{\theta}, t}} $ over the number of particles $ N $, averaged over $ 50 $ simulation runs. We show the ESS of the NPF (in green), the SMC-SMC-UKF (in blue) and the SMC-EnKF-EKF (in red) for $ N = 100 $.

    Figure 4.  Sequences of state values (black line) and estimates (dashed red line) in $ x_1 $ (plot 4a) and $ z_1 $ (plot 4b) over time.

    Figure 5.  Normalized $ {\widehat{\text{ESS}}_{ \boldsymbol{\theta}, t}} $ over the number of particles $ N $, of the first layer of the SMC-EnKF-EKF with $ N = 50 $.

    Figure 6.  Posterior density of the parameters (dashed lines) at time $ \tau = 80 $. The true values are indicated by a black vertical line.

  • [1] A. Abdulle, E. Weinan, B. Engquist and E. Vanden-Eijnden, The heterogeneous multiscale method, Acta Numerica, 21 (2012), 1-87.  doi: 10.1017/S0962492912000025.
    [2] B. D. O. Anderson and J. B. Moore, Optimal Filtering, Englewood Cliffs, 1979.
    [3] C. Andrieu, A. Doucet and R. Holenstein, Particle Markov chain Monte Carlo methods, Journal of the Royal Statistical Society B, 72 (2010), 269-342. 
    [4] H. M. Arnold, Stochastic Parametrisation and Model Uncertainty, PhD thesis, University of Oxford, 2013.
    [5] A. Beskos, D. Crisan, A. Jasra, K. Kamatani and Y. Zhou, A stable particle filter for a class of high-dimensional state-space models, Advances in Applied Probability, 49 (2017), 24-48.  doi: 10.1017/apr.2016.77.
    [6] M. Carlu, F. Ginelli, V. Lucarini and A. Politi, Lyapunov analysis of multiscale dynamics: The slow manifold of the two-scale Lorenz'96 model, preprint, arXiv: 1809.05065, 2018.
    [7] N. Chopin, P. E. Jacob and O. Papaspiliopoulos, SMC$^{2}$: An efficient algorithm for sequential analysis of state space models, Journal of the Royal Statistical Society: Series B (Statistical Methodology), 75 (2013), 397-426.  doi: 10.1111/j.1467-9868.2012.01046.x.
    [8] D. Crisan and J. Míguez, Nested particle filters for online parameter estimation in discrete-time state-space Markov models, Bernoulli, 24 (2018), 3039-3086.  doi: 10.3150/17-BEJ954.
    [9] P. M. Djurić, J. H. Kotecha, J. Zhang, Y. Huang, T. Ghirmai, M. F. Bugallo and and J. Míguez, Particle filtering, IEEE Signal Processing Magazine, 20 (2003), 19-38.  doi: 10.1109/MSP.2003.1236770.
    [10] A. Doucet, S. Godsill and C. Andrieu, On sequential Monte Carlo sampling methods for Bayesian filtering, Statistics and Computing, 10 (2000), 197-208.  doi: 10.1023/A:1008935410038.
    [11] A. Doucet and A. M. Johansen, A tutorial on particle filtering and smoothing: Fifteen years later, Handbook of Nonlinear Filtering, 12 (2009), 656-704. 
    [12] G. Evensen, The ensemble Kalman filter: Theoretical formulation and practical implementation, Ocean Dynamics, 53 (2003), 343-367.  doi: 10.1007/s10236-003-0036-9.
    [13] Z. Fang, A. Gupta and M. Khammash, Effective filtering approach for joint parameter-state estimation in SDEs via Rao-Blackwellization and modularization, preprint, arXiv: 2311.00836, 2023.
    [14] I. Grooms and Y. Lee, A framework for variational data assimilation with superparameterization, Nonlinear Processes in Geophysics, 22 (2015), 601-611.  doi: 10.5194/npg-22-601-2015.
    [15] I. Grooms and G. Robinson, A hybrid particle-ensemble Kalman filter for problems with medium nonlinearity, Plos One, 16 (2021), e0248266.  doi: 10.1371/journal.pone.0248266.
    [16] S. J. Julier and J. Uhlmann, Unscented filtering and nonlinear estimation, Proceedings of the IEEE, 92 (2004), 401-422.  doi: 10.1109/JPROC.2003.823141.
    [17] S. J. Julier, J. K. Uhlmann and H. F. Durrant-Whyte, A new method for the non linear transformation of means and covariances in filters and estimators, IEEE Transactions Automatic Control, 45 (2000), 477-482.  doi: 10.1109/9.847726.
    [18] N. Lingala, N. Sri Namachchivaya, N. Perkowski and H. C. Yeong, Particle filtering in high-dimensional chaotic systems, Chaos: An Interdisciplinary Journal of Nonlinear Science, 22 (2012), 047509.  doi: 10.1063/1.4766595.
    [19] J. S. Liu and R. Chen, Sequential Monte Carlo methods for dynamic systems, Journal of the American Statistical Association, 93 (1998), 1032-1044.  doi: 10.1080/01621459.1998.10473765.
    [20] E. N. Lorenz, Predictability: A problem partly solved, In Proc. Seminar on Predictability, volume 1, 1996.
    [21] J. Mandel, Efficient Implementation of the Ensemble Kalman Filter, University of Colorado at Denver and Health Sciences Center, Center for Computational Mathematics, 2006.
    [22] H. M. Menegaz, J. Y. Ishihara, G. A. Borges and A. N. Vargas, A systematization of the unscented Kalman filter theory, IEEE Transactions on Automatic Control, 60 (2015), 2583-2598.  doi: 10.1109/TAC.2015.2404511.
    [23] J. Míguez, D. Crisan and I. P. Mariño, Particle filtering for Bayesian parameter estimation in a high dimensional state space model, In Proceedings of the 23rd European Signal Processing Conference (EUSIPCO), September 2015.
    [24] C. A. Naesseth, F. Lindsten and T. B. Schön, High-dimensional filtering using nested sequential Monte Carlo, IEEE Transactions on Signal Processing, 67 (2019), 4177-4188.  doi: 10.1109/TSP.2019.2926035.
    [25] G. Pavliotis and A. Stuart, Multiscale Methods: Averaging and Homogenization, Texts Appl. Math., 53, Springer, New York, 2008.
    [26] S. Pérez-Vieites and J. Míguez, Nested Gaussian filters for recursive Bayesian inference and nonlinear tracking in state space models, Signal Processing, 189 (2021), 108295.  doi: 10.1016/j.sigpro.2021.108295.
    [27] S. Pérez-Vieites, I. P. Mariño and J. Míguez, Probabilistic scheme for joint parameter estimation and state prediction in complex dynamical systems, Physical Review E, 98 (2017), 063305.  doi: 10.1103/PhysRevE.98.063305.
    [28] M. Pulido, P. Tandeo, M. Bocquet, A. Carrassi and M. Lucini, Stochastic parameterization identification using ensemble Kalman filtering combined with maximum likelihood methods, Tellus A: Dynamic Meteorology and Oceanography, 70 (2018), 1-17.  doi: 10.1080/16000870.2018.1442099.
    [29] N. Santitissadeekorn and C. Jones, Two-stage filtering for joint state-parameter estimation, Monthly Weather Review, 143 (2015), 2028-2042. 
    [30] Z. Shen, Y. Tang, X. Li, Y. Gao and J. Li, On the localization in strongly coupled ensemble data assimilation using a two-scale Lorenz model, Nonlinear Processes in Geophysics Discussions, (2018), 1-24. 
    [31] T. Tsuyuki, Data assimilation in a two-scale model with Kalman filters, CAS/JSC WGNE Research Activities in Atmospheric and Oceanic Modelling, (2012), 27-28. 
    [32] E. Vanden-Eijnden, Numerical techniques for multi-scale dynamical systems with stochastic effects, Communications in Mathematical Sciences, 1 (2003), 385-391.  doi: 10.4310/CMS.2003.v1.n2.a11.
    [33] E. Vanden-Eijnden, Heterogeneous multiscale methods: A review, Communications in Computational Physics2, (2007), 367-450. 
    [34] G. Vissio and V. Lucarini, A proof of concept for scale-adaptive parametrizations: The case of the Lorenz'96 model, Quarterly Journal of the Royal Meteorological Society, 144 (2018), 63-75.  doi: 10.1002/qj.3184.
    [35] E. Weinan,  Principles of Multiscale Modeling, Cambridge University Press, 2011. 
    [36] E. Weinan and B. Engquist, Multiscale modeling and computation, Notices of the AMS, 50 (2003), 1062-1070. 
    [37] E. Weinan, D. Liu and E. Vanden-Eijnden, Analysis of multiscale methods for stochastic differential equations, Communications on Pure and Applied Mathematics, 58 (2005), 1544-1585.  doi: 10.1002/cpa.20088.
    [38] H. C. Yeong, R. T. Beeson, N. S. Namachchivaya and N. Perkowski, Particle filters with nudging in multiscale chaotic systems: With application to the Lorenz'96 atmospheric model, Journal of Nonlinear Science, (2020), 1-34. 
  • 加载中

Figures(6)

SHARE

Article Metrics

HTML views(3492) PDF downloads(541) Cited by(0)

Access History

Catalog

    /

    DownLoad:  Full-Size Img  PowerPoint
    Return
    Return