Kinetic methods for inverse problems

Michael Herty; Giuseppe Visconti

doi:10.3934/krm.2019042

Article Contents

2019, Volume 12, Issue 5: 1109-1130. Doi: 10.3934/krm.2019042

This issue Previous Article Differentiability in perturbation parameter of measure solutions to perturbed transport equation Next Article On the blow-up criterion and global existence of a nonlinear PDE system in biological transport networks

Kinetic methods for inverse problems

Michael Herty and
Giuseppe Visconti^,

RWTH Aachen University, Institut für Geometrie und Praktische Mathematik, Templergraben 55, 52062 Aachen, Germany

^* Corresponding author: Giuseppe Visconti
^* Corresponding author: Giuseppe Visconti

Received: October 31, 2018

Revised: January 31, 2019

Published: July 2019

Abstract Full Text(HTML) Figure(7) Related Papers Cited by

Abstract

The Ensemble Kalman Filter method can be used as an iterative numerical scheme for parameter identification ornonlinear filtering problems. We study the limit of infinitely large ensemble size and derive the corresponding mean-field limit of the ensemble method. The solution of the inverse problem is provided by the expected value of the distribution of the ensembles and the kinetic equation allows, in simple cases, to analyze stability of these solutions. Further, we present a slight but stable modification of the method which leads to a Fokker-Planck-type kinetic equation. The kinetic methods proposed here are able to solve the problem with a reduced computational complexity in the limit of a large ensemble size. We illustrate the properties and the ability of the kinetic model to provide solution to inverse problems by using examples from the literature.

Keywords:

Mathematics Subject Classification: Primary: 35Q84, 65N21, 93E11; Secondary: 65N75.

Citation:

Full Text(HTML)

Figure 1. Left: vector field of the ODE system (16) with $ (y, G) = (2, 1) $. Red lines are the nullclines. Right: trajectory behavior around the equilibrium $ (\frac{y}{G}, \frac{y^2}{G^2}) = (2, 4) $

Download: Full-size image PowerPoint slide

Figure 2. Left: vector field of the ODE system (24) with $ (y, G) = (2, 1) $. Red lines are the nullclines. Right: trajectory behavior around the equilibrium $ \tilde{F}_1^+ = (2, 8) $

Download: Full-size image PowerPoint slide

Figure 3. Elliptic problem - Test case 1 with $ \gamma = 0.01 $. Top row: plots of the residual $ r $, the projected residual $ R $ and the misfit $ \vartheta $ for $ M = 250,500, 1000 $. Bottom row: plots of the noisy data, the reconstruction of $ p(x) $ and the reconstruction of the control $ u(x) $ at final iteration for $ M = 250,500, 1000 $

Download: Full-size image PowerPoint slide

Figure 4. Elliptic problem - Test case 1 with $ \gamma = 0.1 $. Top row: plots of the residual $ r $, the projected residual $ R $ and the misfit $ \vartheta $ for $ M = 250,500, 1000 $. Bottom row: plots of the noisy data, the reconstruction of $ p(x) $ and the reconstruction of the control $ u(x) $ at final iteration for $ M = 250,500, 1000 $

Download: Full-size image PowerPoint slide

Figure 5. Elliptic problem - Test case 1. Left: spectrum of $ \boldsymbol{{ \mathcal{C}}}(t) G^T \boldsymbol{{\Gamma}} G $ for the initial data with $ \gamma = 0.01 $ and $ \gamma = 0.1 $. Right: adaptive $ \epsilon $ and spectral radius of $ \boldsymbol{{ \mathcal{C}}}(t) G^T \boldsymbol{{\Gamma}} G $ over iterations with $ \gamma = 0.01 $ and $ \gamma = 0.1 $

Download: Full-size image PowerPoint slide

Figure 6. Elliptic problem - Test case 2 with $ \gamma = 0.01 $. Top row: plots of the residual $ r $, the projected residual $ R $ and the misfit $ \vartheta $ for $ J = 25, 25\cdot2^9 $. Middle row: plots of the noisy data and of the reconstruction of $ p(x) $ at final iteration for $ J = 25, 25\cdot2^9 $. Bottom row: plots of the reconstruction of the control $ u(x) $ at final iteration for $ J = 25, 25\cdot2^9 $ and behavior of the relative error $ \frac{\|\overline{\mathbf{{u}}}-\mathbf{{u}}\|_2^2}{\|\overline{\mathbf{{u}}}\|_2^2} $

Download: Full-size image PowerPoint slide

Figure 7. Nonlinear problem. Top row: plots of the density estimation of the initial samples (left) and position of the samples at final iteration (right). Middle row: Marginals of $ u_1 $ (left) and $ u_2 $ (right) as relative frequency plot. Bottom row: residual errors $ r $ and $ R $ (left) and misfit error (right)

Download: Full-size image PowerPoint slide

Related Papers

Cited by

References

[1]	S. I. Aanonsen, G. Nævdal, D. S. Oliver, A. C. Reynolds and B. Vallès, The Ensemble Kalman Filter in Reservoir Engineering–a Review, SPE Journal, 14 (2013), 393-412. doi: 10.2118/117274-PA.
[2]	G. Albi and L. Pareschi, Binary interaction algorithms for the simulation of flocking and swarming dynamics, Multiscale Model. Simul., 11 (2013), 1-29. doi: 10.1137/120868748.
[3]	A. Apte, M. Hairer, A. M. Stuart and J. Voss, Sampling the posterior: An approach to non-Gaussian data assimilation, Phys. D, 230 (2007), 50-64. doi: 10.1016/j.physd.2006.06.009.
[4]	J. O. Berger, Statistical Decision Theory and Bayesian Analysis, 2nd edition, Springer, 1985. doi: 10.1007/978-1-4757-4286-2.
[5]	D. Bianchi, A. Buccini, M. Donatelli and S. Serra-Capizzano, Iterated fractional Tikhonov regularization, Inverse Problems, 31 (2015), 055005, 34pp. doi: 10.1088/0266-5611/31/5/055005.
[6]	D. Bloemker, C. Schillings and P. Wacker, A strongly convergent numerical scheme from ensemble kalman inversion, SIAM J. Numer. Anal., 56 (2018), 2537-2562. doi: 10.1137/17M1132367.
[7]	D. Bloemker, C. Schillings, P. Wacker and S. Weissman, Well Posedness and Convergence Analysis of the Ensemble Kalman Inversion, 2018, Preprint. arXiv: 1810.08463.
[8]	H. Bobovsky and H. Neunzert, On a simulation scheme for the boltzmann equation, Math. Methods Appl. Sci., 8 (1986), 223-233. doi: 10.1002/mma.1670080114.
[9]	M. Burger and F. Lucka, Maximum a posteriori estimates in linear inverse problems with log-concave priors are proper Bayes estimators, Inverse Problems, 30 (2014), 114004, 21pp. doi: 10.1088/0266-5611/30/11/114004.
[10]	R. E. Caflisch, Monte Carlo and quasi-Monte Carlo methods, Acta Numerica, 7 (1998), 1-49. doi: 10.1017/S0962492900002804.
[11]	J. A. Carrillo, M. Fornasier, G. Toscani and F. Vecil, Chapter Particle, kinetic, and hydrodynamic models of swarming, Mathematical Modeling of Collective Behavior in Socio-Economic and Life Sciences, 297–336, Modeling and Simulation in Science, Engineering and Technology, Birkhäuser Boston, 2010. doi: 10.1007/978-0-8176-4946-3_12.
[12]	J. A. Carrillo, L. Pareschi and M. Zanella, Particle based gPC methods for mean-field models of swarming with uncertainty, Commun. Comput. Phys., 25 (2019), 508-531.
[13]	N. K. Chada and X. T. Stuart A. M. Tong, Tikhonov regularization within ensemble Kalman inversion, 2019, arXiv.org/abs/1901.10382.
[14]	E. Cristiani, B. Piccoli and A. Tosin, MS & A: Modeling, Simulation and Applications, vol. 12, chapter Multiscale Modeling of Pedestrian Dynamics, Springer International Publishing, 2014. doi: 10.1007/978-3-319-06620-2.
[15]	M. Dashti and A. M. Stuart, The bayesian approach to inverse problems, Handbook of Uncertainty Quantification, Vol. 1, 2, 3,311–428, Springer, Cham, 2017.
[16]	P. Del Moral, A. Kurtzmann and J. Tugaut, On the stability and the uniform propagation of chaos of a class of Extended Ensemble Kalman-Bucy filters, SIAM J. Control Optim., 55 (2017), 119-155. doi: 10.1137/16M1087497.
[17]	P. Del Moral and J. Tugaut, On the stability and the uniform propagation of chaos properties of Ensemble Kalman-Bucy filters, Ann. Appl. Probab., 28 (2018), 790-850. doi: 10.1214/17-AAP1317.
[18]	L. Desvillettes, On asymptotics of the Boltzmann equation when the collisions become grazing, Transport Theor. Stat., 21 (1992), 259-276. doi: 10.1080/00411459208203923.
[19]	R. J. DiPerna and P. L. Lions, On the Fokker-Planck-Boltzmann equation, Commun. Math. Phys., 120 (1988), 1-23. doi: 10.1007/BF01223204.
[20]	H. W. Engl, M. Hanke and A. Neubauer, Regularization of Inverse Problems, vol. 375, Springer Science and Business Media, 1996.
[21]	O. G. Ernst, B. Sprungk and H.-J. Starkloff, Analysis of the ensemble and polynomial chaos kalman filters in bayesian inverse problems, SIAM/ASA J. Uncertain. Quantif., 3 (2015), 823-851. doi: 10.1137/140981319.
[22]	G. Evensen, Sequential data assimilation with a nonlinear quasi-geostrophic model using monte carlo methods to forecast error statistics, J. Geophys. Res, 99 (1994), 10143-10162. doi: 10.1029/94JC00572.
[23]	G. Evensen, Data Assimilation: The Ensemble Kalman Filter, Springer Verlag, 2009. doi: 10.1007/978-3-642-03711-5.
[24]	M. Fornasier, J. Haskovec and J. Vybíral, Particle systems and kinetic equations modeling interacting agents in high dimension, Multiscale Model. Simul., 9 (2011), 1727-1764. doi: 10.1137/110830617.
[25]	C. W. Groetsch, The Theory of Tikhonov Regularization for Fredholm Equations of the First Kind, vol. 105, Pitman Advanced Publishing Program, 1984.
[26]	S.-Y. Ha and E. Tadmor, From particle to kinetic and hydrodynamic descriptions of flocking, Kinet. Relat. Models, 1 (2008), 415-435. doi: 10.3934/krm.2008.1.415.
[27]	P. C. Hansen, Rank-Deficient and Discrete Ill-Posed Problems: Numerical Aspects of Linear Inversion, SIAM, 1998. doi: 10.1137/1.9780898719697.
[28]	M. Herty and C. Ringhofer, Averaged kinetic models for flows on unstructured networks, Kinet. Relat. Models, 4 (2011), 1081-1096. doi: 10.3934/krm.2011.4.1081.
[29]	M. Iglesias, Iterative regularization for ensemble data assimilation in reservoir models, Computational Geosciences, 19 (2015), 177-212. doi: 10.1007/s10596-014-9456-5.
[30]	M. Iglesias, K. Law and A. M. Stuart, Analysis of the Ensamble Kalman methods for inverse problems, Inverse Problems, 29 (2013), 045001, 20pp. doi: 10.1088/0266-5611/29/4/045001.
[31]	M. Iglesias, K. Law and A. M. Stuart, Evaluation of Gaussian approximations for data assimilation in reservoir models, Comput. Geosci., 17 (2013), 851-885. doi: 10.1007/s10596-013-9359-x.
[32]	R. E. Kalman, A new approach to linear filtering and prediction problems, J. Basic Eng.-T. ASME, 82 (1960), 35-45. doi: 10.1115/1.3662552.
[33]	E. Klann and R. Ramlau, Regularization by fractional filter methods and data smoothing, Inverse Problems, 24 (2008), 0125018, 26pp. doi: 10.1088/0266-5611/24/2/025018.
[34]	E. Kwiatkowski and J. Mandel, Convergence of the square root ensemble Kalman filter in the large ensemble limit, SIAM/ASA J. Uncertain. Quantif., 3 (2015), 1-17. doi: 10.1137/140965363.
[35]	T. Lange and W. Stannat, On the continuous time limit of the ensemble Kalman filter, 2019, arXiv.org/abs/1901.05204.
[36]	K. J. H. Law and A. M. Stuart, Evaluating data assimilation algorithms, Mon. Weather Rev., 140 (2012), 3757-3782. doi: 10.1175/MWR-D-11-00257.1.
[37]	K. J. H. Law, H. Tembine and R. Tempone, Deterministic mean-field ensemble kalman filtering, SIAM J. Sci. Comput., 38 (2016), 1251–1279. doi: 10.1137/140984415.
[38]	F. Le Gland, V. Monbet and V.-D. Tran, Large Sample Asymptotics for the Ensemble Kalman Filter, Research Report RR-7014, INRIA, 2009.
[39]	M. Lemou, Multipole expansions for the Fokker-Planck equation, Numer. Math., 78 (1998), 597-618. doi: 10.1007/s002110050327.
[40]	A. J. Majda and X. T. Tong, Performance of Ensemble Kalman filters in large dimensions, Commun. Pur. Appl. Math., 71 (2018), 892-937. doi: 10.1002/cpa.21722.
[41]	C. Mouhot and L. Pareschi, Fast algorithms for computing the Boltzmann collision operator, Math. Comput., 75 (2006), 1833-1852. doi: 10.1090/S0025-5718-06-01874-6.
[42]	D. S. Oliver, A. C. Reynolds and N. Liu, Inverse Theory for Petroleum Reservoir Characterization and History Matching, Cambridge University Press, 2008. doi: 10.1017/CBO9780511535642.
[43]	L. Pareschi and G. Russo, An introduction to Monte Carlo methods for the Boltzmann equation, in CEMRACS 1999 (Orsay), ESAIM Proc., Soc. Math. Appl. Indust., Paris, 10 (1999), 35–76. doi: 10.1051/proc:2001004.
[44]	L. Pareschi and G. Toscani, Interacting Multiagent Systems. Kinetic equations and Monte Carlo methods, Oxford University Press, 2013.
[45]	L. Pareschi, G. Toscani and C. Villani, Spectral methods for the non cut-off Boltzmann equation and numerical grazing collision limit, Numer. Math., 93 (2003), 527-548. doi: 10.1007/s002110100384.
[46]	C. Schillings and A. M. Stuart, Analysis of the ensamble kalman filter for inverse problems, SIAM J. Numer. Anal., 55 (2017), 1264-1290. doi: 10.1137/16M105959X.
[47]	C. Schillings and A. M. Stuart, Convergence analysis of ensemble Kalman inversion: The linear, noisy case, Applicable Analysis, 97 (2018), 107-123. doi: 10.1080/00036811.2017.1386784.
[48]	A. M. Stuart, Inverse problems: A Bayesian perspective, Acta Numer., 19 (2010), 451-559. doi: 10.1017/S0962492910000061.
[49]	G. Toscani, Kinetic models of opinion formation, Commun. Math. Sci., 4 (2006), 481-496. doi: 10.4310/CMS.2006.v4.n3.a1.
[50]	T. Trimborn, L. Pareschi and M. Frank, Portfolio optimization and model predictive control: A kinetic approach, 2018, Preprint. arXiv: 1711.03291.
[51]	C. Villani, Conservative forms of Boltzmann's collision operator: Landau revisited, ESAIM Math. Model. Numer. Anal., 33 (1999), 209-227. doi: 10.1051/m2an:1999112.