A log-Gaussian Cox process with sequential Monte Carlo for line narrowing in spectroscopy

Teemu Härkönen; Emma Hannula; Matthew T. Moores; Erik M. Vartiainen; Lassi Roininen

doi:10.3934/fods.2023008

Article Contents

2023, Volume 5, Issue 4: 503-519. Doi: 10.3934/fods.2023008

This issue Previous Article Identifiability of interaction kernels in mean-field equations of interacting particles Next Article Hierarchical regularization networks for sparsification based learning on noisy datasets

A log-Gaussian Cox process with sequential Monte Carlo for line narrowing in spectroscopy

1.
School of Engineering Science, LUT University, Yliopistonkatu 34, FI-53850 Lappeenranta, Finland
2.
National Institute for Applied Statistics Research Australia, University of Wollongong, Wollongong NSW 2522, Australia

^*Corresponding author: Teemu Härkönen
^*Corresponding author: Teemu Härkönen

Received: February 2022

Revised: June 2023

Early access: July 2023

Published: December 2023

Abstract / Introduction Full Text(HTML) Figure(5) / Table(1) Related Papers Cited by

Abstract

We propose a statistical model for narrowing line shapes in spectroscopy that are well approximated as linear combinations of Lorentzian or Voigt functions. We introduce a log-Gaussian Cox process to represent the peak locations thereby providing uncertainty quantification for the line narrowing. Bayesian formulation of the method allows for robust and explicit inclusion of prior information as probability distributions for parameters of the model. Estimation of the signal and its parameters is performed using a sequential Monte Carlo algorithm followed by an optimization step to determine the peak locations. Our method is validated using a simulation study and applied to a mineralogical Raman spectrum.

Keywords:

Mathematics Subject Classification: Primary: 62F15, 62L12; Secondary: 78M31.

Citation:

Full Text(HTML)

Figure 1. On top, a spectrum (blue) consisting of $ N = 3 $ Lorentzian line shapes located at locations $ ( l_1, l_2, l_3)^T $ shown in red, yellow, and purple, respectively. Upon successful line narrowing, or deconvolution, we would obtain three individual Dirac delta functions located at $ ( l_1, l_2, l_3)^T $. The aim of this paper is to construct approximate samples for the Dirac delta functions using linear prediction which are further modelled as a log-Gaussian Cox process

Download: Full-size image PowerPoint slide

Figure 2. On top, a summary of the distribution of posterior samples from $ \pi( \mathit{\boldsymbol{x}}_{\text{LN}} \mid \boldsymbol{y}) $ for the spectrum in Figure 1. In the middle, marginalized posterior samples according to Eq. (20) and the corresponding LGCP estimate. At the bottom, the posterior $ \pi( \mathit{\boldsymbol{l}} \mid \mathit{\boldsymbol{z}}) $ for the line shape locations $ \mathit{\boldsymbol{l}} $ obtained by sampling the LGCP local maxima

Download: Full-size image PowerPoint slide

Figure 3. Simulation-based calibration histogram for the rank statistics of the true number of peaks $ N^*_s $. The number of peaks were estimated as the number of local maxima in samples from the LGCP fits, using a GP length scale of $ 0.025 $. The histogram shows a uniform distribution. The solid black line shows the expected value and the shaded gray areas show the 99% confidence intervals

Download: Full-size image PowerPoint slide

Figure 4. At the top, posterior distributions for the line shape parameter $ \gamma $ and the number of line shapes $ N $ along with their respective true parameter values used to generate the synthetic spectrum in red. At the bottom, the corresponding synthetic spectrum (in blue) and the corresponding location posterior $ \pi( \mathit{\boldsymbol{l}} \mid \mathit{\boldsymbol{z}}) $ (in red). Blacks dots denote the locations used to generate the spectrum

Download: Full-size image PowerPoint slide

Figure 5. At the top, posterior distributions for the line shape parameter $ \gamma $ and the number of line shapes $ N $. At the bottom, the observed Raman spectrum of anorthite (in blue) and the corresponding location posterior $ \pi( \mathit{\boldsymbol{l}} \mid \mathit{\boldsymbol{z}}) $ (in red). Blacks dots denote the peak locations found in literature [10]

Download: Full-size image PowerPoint slide

Table 1. Prior distributions for the Lorentz and Voigt line shape parameters $ {\mathit{\boldsymbol{\theta}}} $, the Fourier self-deconvolution cut-off parameter $ M $, and the GP covariance parmeters $ \psi $

Prior	Lorentz	Voigt
$ \pi_0(\gamma) $	$ \mathcal{U}( 1, 30) $	$ \mathcal{U}( 1, 30) $
$ \pi_0(\sigma \mid \gamma) $	NA	$ \mathcal{N}_+( 0.5 \times \gamma, (0.05 \times \gamma)^2 ) $
$ \pi_0(M) $	$ \mathcal{U}( 10, 80) $	$ \mathcal{U}( 10, 80) $
$ \pi_0(\log\{\sigma_\lambda\}) $	$ t( 0,100^2, 10) $	$ t( 0.01,100^2, 10) $

| Show Table

DownLoad: CSV

Related Papers

Cited by

References

[1]	P. J. Brockwell and R. A. Davis, Introduction to Time Series and Forecasting, Springer, 2016. doi: 10.1007/978-3-319-29854-2.
[2]	B. Buttingsrud and B. K. Alsberg, A new maximum entropy-based method for deconvolution of spectra with heteroscedastic noise, Journal of Chemometrics, 18 (2004), 537-547. doi: 10.1002/cem.898.
[3]	N. Chopin and O. Papaspiliopoulos, An Introduction to Sequential Monte Carlo, Springer Series in Statistics. Springer, 2020. doi: 10.1007/978-3-030-47845-2.
[4]	H. Cui, G. Xia, S. Jin, L. Cheng, L. Bai, L. Ma and Y. Fang, Levenberg-Marquardt algorithm with adaptive Tikhonov regularization for bandwidth correction of spectra, Journal of Modern Optics, 67 (2020), 661-670. doi: 10.1080/09500340.2020.1766590.
[5]	H.-D. Dau and N. Chopin, Waste-free sequential Monte Carlo, Journal of the Royal Statistical Society: Series B (Statistical Methodology), 84 (2022), 114-148. doi: 10.1111/rssb.12475.
[6]	P. Del Moral, A. Doucet and A. Jasra, Sequential Monte Carlo samplers, Journal of the Royal Statistical Society: Series B, 68 (2006), 411-436. doi: 10.1111/j.1467-9868.2006.00553.x.
[7]	P. Del Moral, A. Doucet and A. Jasra, An adaptive sequential Monte Carlo method for approximate Bayesian computation, Statistics and Computing, 22 (2012), 1009-1020. doi: 10.1007/s11222-011-9271-y.
[8]	M. Diem, Modern Vibrational Spectroscopy and Micro-Spectroscopy: Theory, Instrumentation and Biomedical Applications, John Wiley & Sons, 2015. doi: 10.1002/9781118824924.
[9]	P. J. Diggle, P. Moraga, B. Rowlingson and B. M. Taylor, Spatial and spatio-temporal log-Gaussian Cox processes: Extending the geostatistical paradigm, Statistical Science, 28 (2013), 542-563. doi: 10.1214/13-STS441.
[10]	J. J. Freeman, A. Wang, K. E. Kuebler, B. L. Jolliff and L. A. Haskin, Characterization of natural feldspars by Raman spectroscopy for future planetary exploration, The Canadian Mineralogist, 46 (2008), 1477-1500. doi: 10.3749/canmin.46.6.1477.
[11]	K. B. Frøhling, T. S. Alstrøm, M. Bache, M. S. Schmidt, M. N. Schmidt, J. Larsen, M. H. Jakobsen and A. Boisen, Surface-enhanced Raman spectroscopic study of DNA and 6-mercapto-1-hexanol interactions using large area mapping, Vibrational Spectroscopy, 86 (2016), 331-336. doi: 10.1016/j.vibspec.2016.08.005.
[12]	S. Gulam Razul, W. J. Fitzgerald and C. Andrieu, Bayesian model selection and parameter estimation of nuclear emission spectra using RJMCMC, Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment, 497 (2003), 492-510. doi: 10.1016/S0168-9002(02)01807-7.
[13]	T. Härkönen, L. Roininen, M. T. Moores and E. M. Vartiainen, Bayesian quantification for coherent anti-Stokes Raman scattering spectroscopy, The Journal of Physical Chemistry B, 124 (2020), 7005-7012. doi: 10.1021/acs.jpcb.0c04378.
[14]	J. K. Kauppinen, D. J. Moffatt, M. R. Hollberg and H. H. Mantsch, A new line-narrowing procedure based on Fourier self-deconvolution, maximum entropy, and linear prediction, Applied Spectroscopy, 45 (1991), 411-416. doi: 10.1366/0003702914337155.
[15]	J. K. Kauppinen, D. J. Moffatt and H. H. Mantsch, Nonlinearity of the maximum entropy method in resolution enhancement, Canadian Journal of Chemistry, 70 (1992), 2887-2894. doi: 10.1139/v92-369.
[16]	J. K. Kauppinen, D. J. Moffatt, H. H. Mantsch and D. G. Cameron, Fourier self-deconvolution: A method for resolving intrinsically overlapped bands, Applied Spectroscopy, 35 (1981), 271-276. doi: 10.1366/0003702814732634.
[17]	J. Kauppinen and J. Partanen, Fourier Transforms in Spectroscopy, Wiley, Berlin, 2001. doi: 10.1002/3527600299.
[18]	B. Lafuente, R. T. Downs, H. Yang and N. Stone, The power of databases: the RRUFF project, In T. Armbruster and R. M. Danisi, editors, Highlights in Mineralogical Crystallography, 1-30. W. De Gruyter, 2015. doi: 10.1515/9783110417104-003.
[19]	G. Last and M. D. Penrose, Lectures on the Poisson Process, Cambridge University Press, 2017. doi: 10.1017/9781316104477.
[20]	D. C. Liu and J. Nocedal, On the limited memory BFGS method for large scale optimization, Mathematical Programming, 45 (1989), 503-528. doi: 10.1007/BF01589116.
[21]	H. Liu, L. Yan, Y. Chang, H. Fang and T. Zhang, Spectral deconvolution and feature extraction with robust adaptive Tikhonov regularization, IEEE Transactions on Instrumentation and Measurement, 62 (2013), 315-327. doi: 10.1109/TIM.2012.2217636.
[22]	V. A. Lórenz-Fonfría and E. Padrós, Maximum entropy deconvolution of infrared spectra: Use of a novel entropy expression without sign restriction, Applied Spectroscopy, 59 (2005), 474-486. doi: 10.1366/0003702053641504.
[23]	J. McLeod and F. Simpson, Validating Gaussian process models with simulation-based calibration, In 2021 IEEE International Conference on Artificial Intelligence Testing (AITest), 101-102. doi: 10.1109/AITEST52744.2021.00028.
[24]	J. Møller, A. R. Syversveen and R. P. Waagepetersen, Log Gaussian Cox processes, Scandinavian Journal of Statistics, 25 (1998), 451-482. doi: 10.1111/1467-9469.00115.
[25]	M. T. Moores, J. Carson, K. Gracie, K. Faulds, D. Graham and M. Girolami, Bayesian modelling and quantification of Raman spectroscopy, 2016, arXiv preprint, arXiv: 1604.07299.
[26]	J. G. Rasmussen, Bayesian inference for Hawkes processes, Methodology and Computing in Applied Probability, 15 (2011), 623-642. doi: 10.1007/s11009-011-9272-5.
[27]	C. Ritter, Statistical analysis of spectra from electron spectroscopy for chemical analysis, The Statistician, 43 (1994), 111-127. doi: 10.2307/2348937.
[28]	S. Särkkä, Bayesian Filtering and Smoothing, Cambridge University Press, 2013. doi: 10.1017/CBO9781139344203.
[29]	L. Serra, M. Saez, J. Mateu, D. Varga, P. Juan, C. Díaz-Ávalos and H. Rue, Spatio-temporal log-Gaussian Cox processes for modelling wildfire occurrence: The case of Catalonia, 1994-2008, Environmental and Ecological Statistics, 21 (2014), 531-563. doi: 10.1007/s10651-013-0267-y.
[30]	S. Shirota and A. E. Gelfand, Space and circular time log Gaussian Cox processes with application to crime event data, The Annals of Applied Statistics, 11 (2017), 481-503. doi: 10.1214/16-AOAS960.
[31]	E. Smith and G. Dent, Modern Raman Spectroscopy: A Practical Approach, 2$^nd$ edition, John Wiley & Sons, 2019. doi: 10.1002/9781119440598.
[32]	R. H. Stewart, Introduction to Physical Oceanography, Texas A & M University, 2004.
[33]	J. Suuronen, M. Emzir, S. Lasanen, S. Särkkä and L. Roininen, Roininen, Enhancing industrial X-ray tomography by data-centric statistical methods, Data-Centric Engineering, 1 (2020), e10. doi: 10.1017/dce.2020.10.
[34]	S. Talts, M. Betancourt, D. Simpson, A. Vehtari and A. Gelman, Validating Bayesian inference algorithms with simulation-based calibration, 2020., arXiv preprint, arXiv: 1804.06788.
[35]	J. Vanhatalo, J. Riihimäki, J. Hartikainen, P. Jylänki, V. Tolvanen and A. Vehtari, GPstuff: Bayesian modeling with Gaussian processes, Journal of Machine Learning Research, 14 (2013), 1175-1179.
[36]	I. I. Virtanen, H. W. Tesfaw, L. Roininen, S. Lasanen and A. Aikio, Bayesian filtering in incoherent scatter plasma parameter fits, Journal of Geophysical Research: Space Physics, 126 (2021), e2020JA028700. doi: 10.1029/2020JA028700.
[37]	P. Whittle, Estimation and information in stationary time series, Arkiv för Matematik, 2 (1953), 423-434. doi: 10.1007/BF02590998.