On periodic parameter identification in stochastic differential equations

Pingping Niu; Shuai Lu; Jin Cheng

doi:10.3934/ipi.2019025

Article Contents

2019, Volume 13, Issue 3: 513-543. Doi: 10.3934/ipi.2019025

This issue Previous Article A stochastic approach to reconstruction of faults in elastic half space Next Article Inverse obstacle scattering for elastic waves in three dimensions

On periodic parameter identification in stochastic differential equations

Shanghai Key Laboratory for Contemporary Applied Mathematics, Key Laboratory of Mathematics for Nonlinear Sciences and School of Mathematical Sciences, Fudan University, Shanghai 200433, China

^† Corresponding author: Shuai Lu
^† Corresponding author: Shuai Lu

Received: March 31, 2018

Revised: September 30, 2018

Published: March 2019

S. Lu is supported by National Key Research and Development Program of China (No. 2017YFC1404103), NSFC (No.91730304, 11522108), Shanghai Municipal Education Commission (No.16SG01) and Special Funds for Major State Basic Research Projects of China (2015CB856003). J. Cheng is supported by NSFC (key projects no.11331004, no.11421110002) and the Programme of Introducing Talents of Discipline to Universities (number B08018).

Abstract Full Text(HTML) Figure(7) / Table(3) Related Papers Cited by

Abstract

Periodic parameters are common and important in stochastic differential equations (SDEs) arising in many contemporary scientific and engineering fields involving dynamical processes. These parameters include the damping coefficient, the volatility or diffusion coefficient and possibly an external force. Identification of these periodic parameters allows a better understanding of the dynamical processes and their hidden intermittent instability. Conventional approaches usually assume that one of the parameters is known and focus on the recovery of rest parameters. By introducing the decorrelation time and calculating the standard Gaussian statistics (mean, variance) explicitly for the scalar Langevin equations with periodic parameters, we propose a parameter identification approach to simultaneously recovering all these parameters by observing a single trajectory of SDEs. Such an approach is summarized in form of regularization schemes with noisy operators and noisy right-hand sides and is further extended to parameter identification of SDEs which are indirectly observed by other random processes. Numerical examples show that our approach performs well in stable and weakly unstable regimes but may fail in strongly unstable regime which is induced by the strong intermittent instability itself.

Keywords:

Mathematics Subject Classification: Primary: 65L09; Secondary: 60H10.

Citation:

Full Text(HTML)

Figure 1. Pathwise solutions of the stochastic differential equation (4) with different parameters in Table 1. Upper (stable regime); middle (weakly unstable regime) and bottom (strongly unstable regime). Each panel presents a segment of $ v(t) $ for $ t\in [100,102] $ whereas the long path of the solution $ v(t) $ for $ t\in [0,20000] $ is presented in the small picture in each panel

Download: Full-size image PowerPoint slide

Figure 2. Empirical values of $ 5 $ trajectories of the random process $ v(t) $ in (4) with different parameters in Table 1. Upper row (stable regime); middle row (weakly unstable regime) and bottom row (strongly unstable regime). The red solid line in each panel is the exact value. The blue dashed lines are empirical values of $ 5 $ different trajectories

Download: Full-size image PowerPoint slide

Figure 3. Pathwise solutions of the stochastic differential equations (23)-(24) with different parameters in Table 1 and (30). Upper row: $ v(t) $ and $ u(t) $ (stable regime); middle row: $ v(t) $ and $ u(t) $ (weakly unstable regime) and bottom row: $ v(t) $ and $ u(t) $ (strongly unstable regime). Each panel presents a segment of $ v(t) $ or $ u(t) $ for $ t\in [100,102] $ whereas the long path of the solutions for $ t\in [0,40000] $ is presented in the small picture in each panel

Download: Full-size image PowerPoint slide

Figure 4. Empirical values of $ 5 $ trajectories of the random process $ u(t) $ in (23)-(24) with different parameters in Table 1 and (30). Upper row (stable regime); middle row (weakly unstable regime) and bottom row (strongly unstable regime). The red solid line in each panel is the exact value. The blue dashed lines are empirical values of $ 5 $ different trajectories

Download: Full-size image PowerPoint slide

Figure 5. Parameter identification approach for direct observation $ v(t) $ in (4). Upper row: exact and reconstructed $ \gamma_v $, $ f_{v} $ and $ \sigma_v^2 $ (stable regime); middle row: exact and reconstructed $ \gamma_v $, $ f_{v} $ and $ \sigma_v^2 $ (weakly unstable regime) and bottom row: exact and reconstructed $ \gamma_v $, $ f_{v} $ and $ \sigma_v^2 $ (strongly unstable regime). Small figures in each panel are the exact (red solid line) and empirical (blue dashed line) decorrelation time, mean and variance of $ v(t) $ which are used in the parameter identification approach (20)

Download: Full-size image PowerPoint slide

Figure 6. Parameter identification approach for indirect observation $ u(t) $ in (23)-(24). Upper row: exact and reconstructed $ \gamma_v $, $ f_{v} $ and $ \sigma_v^2 $ (stable regime); middle row: exact and reconstructed $ \gamma_v $, $ f_{v} $ and $ \sigma_v^2 $ (weakly unstable regime) and bottom row: exact and reconstructed $ \gamma_v $, $ f_{v} $ and $ \sigma_v^2 $ (strongly unstable regime). Small figures in each panel are the exact (red solid line) and empirical (blue dashed line) decorrelation time, mean and variance of $ u(t) $ which are used in the parameter identification approach (28)

Download: Full-size image PowerPoint slide

Figure 7. Empirical values of $ 5 $ trajectories of the random process $ v(t) $ (upper row) and $ u(t) $ (bottom row) with reconstructed parameters of the strongly unstable regime. The red solid line in each panel is the exact Gaussian statistics. The blue dashed lines are empirical Gaussian statistics of $ 5 $ different trajectories by the reconstructed parameters

Download: Full-size image PowerPoint slide

Table 1. Parameters of the stochastic differential equations (4) and (23)

Parameters	Value
Damping parameter	(stable regime): $ \gamma_v(\zeta)=2\sin(2\pi\zeta+\pi)+2.05 $ (weakly unstable regime): $ \gamma_v(\zeta)=2\sin(2\pi\zeta+\pi)+1.9 $ (strongly unstable regime): $ \gamma_v(\zeta)=2\sin(2\pi\zeta+\pi)+0.05 $
Force parameter	$ f_v(\zeta)=0.1\sin(4\pi\zeta)+0.2 $
Volatility parameter	$ \sigma_v^2(\zeta)=\left(0.1\sin(2\pi\zeta)+0.3\right)^2 $.

| Show Table

DownLoad: CSV

Table 2. Direct observation. Columns 2-4: $ L^2- $relative errors of exact and empirical Gaussian statistics of $ v(t) $. Final three columns: $ L^2- $relative errors of the exact and reconstructed parameters. The observation time is $ [15000,20000] $

	Decorrelation time	Mean	Variance	$ \gamma_v $	$ f_v $	$ \sigma_v^2 $
stable	0.0287	0.0170	0.0173	0.0347	0.0706	0.0271
weakly unstable	0.0347	0.0295	0.0173	0.0601	0.0949	0.0662
strongly unstable	0.0049	0.0274	0.0310	0.0178	0.5762	0.3591

| Show Table

DownLoad: CSV

Table 3. Indirect observation. Columns 2-4: $ L^2- $relative errors of exact and empirical Gaussian statistics of $ u(t) $. Final three columns: $ L^2- $relative errors of the exact and reconstructed parameters. The observation time is $ [30000,40000] $

	Decorrelation time	Mean	Variance	$ \gamma_v $	$ f_v $	$ \sigma_v^2 $
stable	0.0108	0.0043	0.0098	0.0496	0.0916	0.0613
weakly unstable	0.0066	0.0099	0.0151	0.0208	0.1724	0.0773
strongly unstable	0.0090	0.0142	0.0575	0.0187	0.3866	0.1462

| Show Table

DownLoad: CSV

Related Papers

Cited by

References

[1]	G. Bao, C. Chen and P. Li, Inverse random source scattering problems in several dimensions, SIAM/ASA Journal on Uncertainty Quantification, 4 (2016), 1263-1287. doi: 10.1137/16M1067470.
[2]	G. Bao, S. N. Chow and H. Zhou, An inverse random source problem for the Helmholtz equation, Mathematics of Computation, 83 (2014), 215-233. doi: 10.1090/S0025-5718-2013-02730-5.
[3]	G. Bao and X. Xu, An inverse random source problem in quantifying the elastic modulus of nanomaterials, Inverse Problems, 29 (2013), 015006, 16 pp. doi: 10.1088/0266-5611/29/1/015006.
[4]	G. Bao and X. Xu, Identification of the material properties in nonuniform nanostructures, Inverse Problems, 31 (2015), 125003, 11 pp. doi: 10.1088/0266-5611/31/12/125003.
[5]	M. Branicki and A. J. Majda, Quantifying Bayesian filter performance for turbulent dynamical systems through information theory, Commun. Math. Sci., 12 (2014), 901-978. doi: 10.4310/CMS.2014.v12.n5.a6.
[6]	D. G. Cacuci, I. M. Navon and M. Ionescu-Bujor, Computational Methods for Data Evaluation and Assimilation, CRC Press, Boca Raton, FL, 2014.
[7]	N. Chen, D. Giannakis, R. Herbei and A. J. Majda, An MCMC algorithm for parameter estimation in signals with hidden intermittent instability, SIAM/ASA Journal on Uncertainty Quantification, 2 (2014), 647-669. doi: 10.1137/130944977.
[8]	M. Cristofol and L. Roques, Simultaneous determination of the drift and diffusion coefficients in stochastic differential equations, Inverse Problems, 33 (2017), 095006, 12 pp. doi: 10.1088/1361-6420/aa7a1c.
[9]	F. Dunker and T. Hohage, On parameter identification in stochastic differential equations by penalized maximum likelihood, Inverse Problems, 30 (2014), 095001, 20 pp. doi: 10.1088/0266-5611/30/9/095001.
[10]	H. W. Engl, M. Hanke and A. Neubauer, Regularization of Inverse Problems, Mathematics and its Applications, 375. Kluwer Academic Publishers Group, Dordrecht, 1996. viii+321 pp.
[11]	H. Gao, K. Wang, F. Wei and X. Ding, Massera-type theorem and asymptotically periodic logistic equations, Nonlinear Analysis: Real World Applications, 7 (2006), 1268-1283. doi: 10.1016/j.nonrwa.2005.11.008.
[12]	B. Gershgorin, J. Harlim and A. J. Majda, Test models for improving filtering with model errors through stochastic parameter estimation, Journal of Computational Physics, 229 (2010), 1-31. doi: 10.1016/j.jcp.2009.08.019.
[13]	A. Golightly and D. J. Wilkinson, Bayesian inference for stochastic kinetic models using a diffusion approximation, Biometrics, 61 (2005), 781-788. doi: 10.1111/j.1541-0420.2005.00345.x.
[14]	B. Kaltenbacher and B. Pedretscher, Parameter estimation in SDEs via the Fokker-Planck equation: Likelihood function and adjoint based gradient computation, Journal of Mathematical Analysis and Applications, 465 (2018), 872-884. doi: 10.1016/j.jmaa.2018.05.048.
[15]	W. J. Lee and A. Stuart, Derivation and analysis of simplified filters, Communications in Mathematical Sciences, 15 (2017), 413-450. doi: 10.4310/CMS.2017.v15.n2.a6.
[16]	P. Li, An inverse random source scattering problem in inhomogeneous media, Inverse Problems, 27 (2011), 035004, 22 pp. doi: 10.1088/0266-5611/27/3/035004.
[17]	S. Lu and S. V. Pereverzev, Regularization Theory for Ill-Posed Problems: Selected Topics, Inverse and Ill-posed Problems Series, 58 De Gruyter, Berlin, 2013. doi: 10.1515/9783110286496.
[18]	S. Lu, S. V. Pereverzev, Y. Shao and U. Tautenhahn, On the generalized discrepancy principle for Tikhonov regularization in Hilbert scales, Journal of Integral Equations and Application, 22 (2010), 483-517. doi: 10.1216/JIE-2010-22-3-483.
[19]	A. J. Majda and J. Harlim, Filtering Complex Turbulent Systems, Cambridge University Press, Cambridge, 2012. x+357 pp. doi: 10.1017/CBO9781139061308.
[20]	A. J. Majda and X. Wang, Non-linear Dynamics and Statistical Theories for Basic Geophysical Flows, Cambridge University Press, Cambridge, 2006. doi: 10.1017/CBO9780511616778.
[21]	B. Øksendal, Stochastic Differential Equations, 6$^{th}$ ed., Springer, Heidelberg, 2003. doi: 10.1007/978-3-642-14394-6.
[22]	O. Papaspiliopoulos, Y Pokern, G. O. Roberts and A. M. Stuart, Nonparametric estimation of diffusions: A differential equations approach, Biometrika, 99 (2012), 511-531. doi: 10.1093/biomet/ass034.
[23]	G. O. Roberts and O. Stramer, On inference for partially observed nonlinear diffusion model using Metropolis–Hastings algorithms, Biometrika, 88 (2001), 603-621. doi: 10.1093/biomet/88.3.603.
[24]	H. Sorensen, Parametric inference for diffusion processes observed at discrete points in time: A survey, International Statistical Review, 72 (2004), 337-354. doi: 10.1111/j.1751-5823.2004.tb00241.x.
[25]	O. Stramer and M. Bognar, Bayesian inference for irreducible diffusion processes using the pseudo-marginal approach, Bayesian Analysis, 6 (2011), 231-258. doi: 10.1214/11-BA608.
[26]	A. M. Stuart, Inverse problems: A Bayesian perspective, Acta Numer., 19 (2010), 451-559. doi: 10.1017/S0962492910000061.
[27]	U. Tautenhahn, Regularization of linear ill-posed problems with noisy right hand side and noisy operator, J. Inv. Ill-Posed Problems, 16 (2008), 507-523. doi: 10.1515/JIIP.2008.027.
[28]	C. R. Vogel, Computational Methods for Inverse Problems, With a foreword by H. T. Banks. Frontiers in Applied Mathematics, 23. Society for Industrial and Applied Mathematics (SIAM), Philadelphia, PA, 2002. doi: 10.1137/1.9780898717570.