This paper is devoted to the problem of sampling Gaussian distributions in high dimension. Solutions exist for two specific structures inverse covariance: sparse and circulant. The proposed algorithm valid a more general case especially as it emerges linear problems well some hierarchical or latent models. It relies on perturbation-optimization principle: adequate stochastic perturbation criterion optimization perturbed criterion. proved that optimizer sample target distribution. main...

We deal with an electromagnetic inverse scattering problem where the goal is to characterize unknown objects from measurements of scattered fields that result their interaction a known interrogating wave in microwave frequency range. This nonlinear and ill-posed tackled experimental data collected laboratory-controlled experiment led at Institut Fresnel (Marseille, France), which consist time-harmonic electric field values measured several discrete frequencies. The modelling wave–object...

We study price formation in intraday electricity markets the presence of intermittent renewable generation. consider setting where a major producer may interact strategically with large number small producers. Using stochastic control theory, we identify optimal strategies agents market impact and exhibit Nash equilibrium closed form asymptotic framework mean field games player.

Uncertainty quantification of predictive models is crucial in decision-making problems. Conformal prediction a general and theoretically sound answer. However, it requires exchangeable data, excluding time series. While recent works tackled this issue, we argue that Adaptive Inference (ACI, Gibbs Cand{\`e}s, 2021), developed for distribution-shift series, good procedure series with dependency. We analyse the impact learning rate on its efficiency auto-regressive case. propose parameter-free...

Abstract Change points detection in time series is an important area of research statistics, has a long history and many applications. However, very often change point analysis only focused on the changes mean value some quantity process. In this work we consider with discrete which may contain finite number probability density functions (pdf). We focus case where data all segments are modeled by Gaussian different means, variances correlation lengths. put prior law occurances (Poisson...

This paper deals with Gibbs samplers that include high dimensional conditional Gaussian distributions. It proposes an efficient algorithm avoids the sampling and relies on a random excursion along small set of directions. The is proved to converge, i.e., drawn samples are asymptotically distributed according target distribution. Our main motivation in inverse problems related general linear observation models their solution hierarchical Bayesian framework implemented through algorithms....

A Fourier transform IR spectrometer and a mass have been coupled to pyrocarbon low pressure CVD reactor in order investigate the gas phase during pyrolysis of propane range 600–1060°C. The variations amounts intermediate species such as ethylene, acetylene, benzene are followed by infrared absorption measurements function temperature residence time. Mass spectrometry allows additional minor be identified. overall study permits reaction pathway from diacetylene, benzene, heavier proposed.

In this paper we propose a stochastic algorithm applied to an electromagnetic inverse scattering problem. The objective is characterise unknown object from measurements of the scattered fields at different frequencies and for several illuminati

In order to develop a computer-assisted process optimization of 1- x Ga As y P metalorganic chemical vapor deposition (MOCVD), the kinetics GaAs growth was studied as first step. For accumulation reaction data source materials, decomposition rates trimethylgallium (TMGa) and tertiarybutylarsine (TBAs) were using flow tube reactor Fourier transform infrared spectrometer (FT-IR). Special attention paid effect TBAs concentration on TMGa. The rate profile in commercial MOCVD analyzed through...

We develop a tractable equilibrium model for price formation in intraday electricity markets the presence of intermittent renewable generation. Using stochastic control theory we identify optimal strategies agents with market impact and exhibit Nash closed form finite number as well asymptotic framework mean field games. Our reproduces empirical features prices, such increasing volatility at approach delivery date correlation between infeed forecasts, relates these characteristics like...

The paper deals with Gibbs samplers that include high-dimensional conditional Gaussian distributions. It proposes an efficient algorithm only requires a scalar sampling. relies on random excursion along direction. is proved to converge, i.e. the drawn samples are asymptotically under target distribution. Our original motivation in unsupervised inverse problems related general linear observation models and their solution hierarchical Bayesian framework implemented through sampling algorithms....

In this work we propose a Bayesian framework for data fusion of multivariate signals which arises in imaging systems. More specifically, consider the case where have observed two images same object through different processes. The objective is then to coherent approach combine these sets obtain segmented image can be considered as result images. proposed based on Hidden Markov Modeling (HMM) with common segmentation, or equivalently, hidden classification label variables modeled by Potts...

Abstract We statistically analyze a multivariate Heath‐Jarrow‐Morton diffusion model with stochastic volatility. The volatility process of the first factor is left totally unspecified while second product an unknown and exponential function time to maturity. This term includes some real parameter measuring rate increase as goes From historical data, we efficiently estimate maturity in sense constructing estimator that achieves optimal information bound semiparametric setting. also...

In this work we consider time series with a finite number of discrete point changes. We assume that the data in each segment follows different probability density functions (pdf). focus on case where all segments are modeled by Gaussian means, variances and correlation lengths. put prior law change instances (Poisson process) as well these parameters(conjugate priors) give expression posterior probality distributions points. The computations done using an appropriate Markov Chain Monte Carlo...