Publications

2022

• Spatio-temporal mixture process estimation to detect population dynamical changes
  
  • Pruilh Solange
  • Jannot Anne-Sophie
  • Allassonnière Stéphanie
  Artificial Intelligence in Medicine, Elsevier, 2022, 126, pp.102258. Population monitoring is a challenge in many areas such as public health or ecology. We propose a method to model and monitor population distributions over space and time, in order to build an alert system for spatio-temporal data evolution. Assuming that mixture models can correctly model populations, we propose new versions of the Expectation-Maximization algorithm to better estimate both the number of clusters together with their parameters. We then combine these algorithms with a temporal statistical model, allowing to detect dynamical changes in population distributions, and name it a spatio-temporal mixture process (STMP). We test STMP on synthetic data, and consider several different behaviors of the distributions, to adjust this process. Finally, we validate STMP on a real data set of positive diagnosed patients to corona virus disease 2019. We show that our pipeline correctly models evolving real data and detects epidemic changes. (10.1016/j.artmed.2022.102258)
  DOI : 10.1016/j.artmed.2022.102258
• Fixed-distance multipoint formulas for the scattering amplitude from phaseless measurements
  
  • Novikov Roman
  • Sivkin Vladimir
  Inverse Problems, IOP Publishing, 2022, 38 (2), pp.025012. We give new formulas for finding the complex (phased) scattering amplitude at fixed frequency and angles from absolute values of the scattering wave function at several points $x_1,..., x_m$. In dimension $d\geq 2$, for $m>2$, we significantly improve previous results in the following two respects. First, geometrical constraints on the points needed in previous results are significantly simplified. Essentially, the measurement points $x_j$ are assumed to be on a ray from the origin with fixed distance $\tau=|x_{j+1}- x_j|$, and high order convergence (linearly related to $m$) is achieved as the points move to infinity with fixed $\tau$. Second, our new asymptotic reconstruction formulas are significantly simpler than previous ones. In particular, we continue studies going back to [Novikov, Bull. Sci. Math. 139(8), 923-936, 2015]. (10.1088/1361-6420/ac44db)
  DOI : 10.1088/1361-6420/ac44db
• Weak Langmuir turbulence in disordered multimode optical fibers
  
  • Baudin Kilian
  • Garnier Josselin
  • Fusaro Adrien
  • Berti Nicolas
  • Millot Guy
  • Picozzi Antonio
  Physical Review A, American Physical Society, 2022, 105 (1), pp.013528. We consider the propagation of temporally incoherent waves in multimode optical fibers (MMFs) in the framework of the multimode nonlinear Schrödinger (NLS) equation accounting for the impact of the natural structural disorder that affects light propagation in standard MMFs (random mode coupling and polarization fluctuations). By averaging the dynamics over the fast disordered fluctuations, we derive a Manakov equation from the multimode NLS equation, which reveals that the Raman effect introduces a previously unrecognized nonlinear coupling among the modes. Applying the wave turbulence theory on the Manakov equation, we derive a very simple scalar kinetic equation describing the evolution of the multimode incoherent waves. The structure of the kinetic equation is analogous to that developed in plasma physics to describe weak Langmuir turbulence. The extreme simplicity of the derived kinetic equation provides physical insight into the multimode incoherent wave dynamics. It reveals the existence of different collective behaviors where all modes self-consistently form a multimode spectral incoherent soliton state. Such an incoherent soliton can exhibit a discrete behavior characterized by collective synchronized spectral oscillations in frequency space. The theory is validated by accurate numerical simulations: The simulations of the generalized multimode NLS equation are found in quantitative agreement with those of the derived scalar kinetic equation without using adjustable parameters. (10.1103/PhysRevA.105.013528)
  DOI : 10.1103/PhysRevA.105.013528
• Firm non-expansive mappings in weak metric spaces
  
  • Gutiérrez Armando W.
  • Walsh Cormac
  Archiv der Mathematik, Springer Verlag, 2022,  119, pp.389-400. We introduce the notion of firm non-expansive mapping in weak metric spaces, extending previous work for Banach spaces and certain geodesic spaces. We prove that, for firm non-expansive mappings, the minimal displacement, the linear rate of escape, and the asymptotic step size are all equal. This generalises a theorem by Reich and Shafrir.
• Numerical reconstruction from the Fourier transform on the ball using prolate spheroidal wave functions
  
  • Isaev Mikhail
  • Novikov Roman
  • Sabinin Grigory
  Inverse Problems, IOP Publishing, 2022. We implement numerically formulas of [Isaev, Novikov, arXiv:2107.07882, hal-03289374] for finding a compactly supported function v on R^d , d ≥ 1, from its Fourier transform F[v] given within the ball B_r. For the one-dimensional case, these formulas are based on the theory of prolate spheroidal wave functions, which arise, in particular, in the singular value decomposition of the aforementioned band-limited Fourier transform for d = 1. In multidimensions, these formulas also include inversion of the Radon transform. In particular, we give numerical examples of super-resolution, that is, recovering details beyond the diffraction limit. (10.1088/1361-6420/ac87cb)
  DOI : 10.1088/1361-6420/ac87cb
• EV-GAN: Simulation of extreme events with ReLU neural networks
  
  • Allouche Michaël
  • Girard Stéphane
  • Gobet Emmanuel
  Journal of Machine Learning Research, Microtome Publishing, 2022, 23 (150), pp.1--39. Feedforward neural networks based on Rectified linear units (ReLU) cannot efficiently approximate quantile functions which are not bounded, especially in the case of heavy-tailed distributions. We thus propose a new parametrization for the generator of a Generative adversarial network (GAN) adapted to this framework, basing on extreme-value theory. An analysis of the uniform error between the extreme quantile and its GAN approximation is provided: We establish that the rate of convergence of the error is mainly driven by the second-order parameter of the data distribution. The above results are illustrated on simulated data and real financial data. It appears that our approach outperforms the classical GAN in a wide range of situations including high-dimensional and dependent data. The code is available at \url{https://github.com/michael-allouche/extreme-value-GAN}.
• Asymptotic Analysis of a Matrix Latent Decomposition Model
  
  • Mantoux Clément
  • Durrleman​ Stanley
  • Allassonnière Stéphanie
  ESAIM: Probability and Statistics, EDP Sciences, 2022, 26, pp.208-242. Matrix data sets arise in network analysis for medical applications, where each network belongs to a subject and represents a measurable phenotype. These large dimensional data are often modeled using lower-dimensional latent variables, which explain most of the observed variability and can be used for predictive purposes. In this paper, we provide asymptotic convergence guarantees for the estimation of a hierarchical statistical model for matrix data sets. It captures the variability of matrices by modeling a truncation of their eigendecomposition. We show that this model is identifiable, and that consistent Maximum A Posteriori (MAP) estimation can be performed to estimate the distribution of eigenvalues and eigenvectors. The MAP estimator is shown to be asymptotically normal for a restricted version of the model. (10.1051/ps/2022004)
  DOI : 10.1051/ps/2022004
• Leveraging Deep Learning for Efficient Explicit MPC of High-Dimensional and Non-linear Chemical Processes
  
  • Shokry Ahmed
  • El Qassime Mehdi Abou
  • Moulines Eric
  , 2022, 51, pp.1171-1176. (10.1016/B978-0-323-95879-0.50196-X)
  DOI : 10.1016/B978-0-323-95879-0.50196-X