Assessing the intensity of a epidemic, such as the COVID19 pandemic, during the epidemic outbreak, constitutes a significant technical challenge with high societal stakes. Elaborating on classical epidemiological models, this work aims to define a hierarchical Bayesian model that permits the robust estimation of the temporal evolution of the pandemic intensity despite highly corrupted daily new infection counts. It also outputs uncertainty assessment, in the form of credibility intervals robust to the priors choice, accounting for uncertainties on model parameters. The estimation is performed by carefully designed Monte Carlo samplers. The relevance of the proposed estimation procedure is illustrated on real COVID-19 pandemic data for several countries and periods, made available from the Johns Hopkins University repository.

Sampling from a distribution on the real d-space, whose density is nonsmooth and log-concave, is a computational issue that often arises in Machine Learning and Statistics. Langevin-based Hastings-Metropolis methods were proposed: they extend the Unadjusted Langevin Algorithm by using proximal methods to define a smoothed version of the density of interest. We consider the case when these extensions do not apply: the involved proximal operators do not have closed-form expressions and the density is defined on a subset of the real d-space. We derive new Gaussian proposal mechanisms in a Metropolis Adjusted Langevin Algorithm, which use first-order information about the density function. We numerically compare these strategies and discuss the benefits of a change of geometry. The gain in using partial updates of the parameter instead of global updates is also illustrated.

The Expectation Maximization (EM) algorithm is a key reference for inference in latent variable models; unfortunately, its computational cost is prohibitive in the large scale learning setting. In this paper, we propose an extension of the Stochastic Path-Integrated Differential EstimatoR EM (SPIDER-EM) and derive complexity bounds for this novel algorithm, designed to solve smooth nonconvex finite-sum optimization problems. We show that it reaches the same state of the art complexity bounds as SPIDER-EM; and provide conditions for a linear rate of convergence. Numerical results support our findings.

The Expectation Maximization (EM) algorithm is the default algorithm for inference in latent variable models. As in any other field of machine learning, applications of latent variable models to very large datasets make the use of advanced parallel and distributed architecture mandatory. This paper introduces FedEM, which is the first extension of the EM algorithm to the federated learning context. FedEM is a new communication efficient method, which handles partial participation of local devices, and is robust to heterogeneous distribution of the datasets. To alleviate the communication bottleneck, FedEM compresses appropriately defined complete data sufficient statistics. We also develop and analyze an extension of FedEM to further incorporate a variance reduction scheme. In all cases, we derive finite-time complexity bounds for smooth non-convex problems. Numerical results are presented to support our theoretical findings, as well as an application to federated missing values imputation for biodiversity monitoring.

The Expectation Maximization (EM) algorithm is of key importance for inference in latent variable models including mixture of regressors and experts, missing observations. This paper introduces a novel EM algorithm, called {\tt SPIDER-EM}, for inference from a training set of size $n$, $n \gg 1$. At the core of our algorithm is an estimator of the full conditional expectation in the {\sf E}-step, adapted from the stochastic path integral differential estimator ({\tt SPIDER}) technique. We derive finite-time complexity bounds for smooth non-convex likelihood: we show that for convergence to an $\epsilon$-approximate stationary point, the complexity scales as $K_{Opt} (n,\epsilon )={\cal O}(\epsilon^{-1})$ and $K_{CE}( n,\epsilon ) = n+ \sqrt{n} {\cal O}( \epsilon^{-1} )$, where $K_{Opt}( n,\epsilon )$ and $K_{CE}(n, \epsilon )$ are respectively the number of {\sf M}-steps and the number of per-sample conditional expectations evaluations. This improves over the state-of-the-art algorithms. Numerical results support our findings.

We study a version of the proximal gradient algorithm for which the gradient is intractable and is approximated by Monte Carlo methods (and in particular Markov Chain Monte Carlo). We derive conditions on the step size and the Monte Carlo batch size under which convergence is guaranteed: both increasing batch size and constant batch size are considered. We also derive non- asymptotic bounds for an averaged version. Our results cover both the cases of biased and unbiased Monte Carlo approximation. To support our findings, we discuss the inference of a sparse generalized linear model with random effect and the problem of learning the edge structure and parameters of sparse undirected graphical models.

Coverage optimization is an important process for the operator, as it is a crucial prerequisite toward offering a satisfactory quality of service to the end users. The first step of this process is coverage prediction, which can be performed by interpolating geo-located measurements reported to the network by mobile user's equipments. In the previous works, we proposed a low complexity coverage prediction algorithm based on the adaptation of the geo-statistics fixed rank kriging (FRK) algorithm. We supposed that the geo-location information reported with the radio measurements was perfect, which is not the case in reality. In this paper, we study the impact of location uncertainty on the coverage prediction accuracy and we extend the previously proposed algorithm to include geo-location error in the prediction model. We validate the proposed algorithm using both simulated and real-field measurements. The FRK is extended to take into account that the location uncertainty proves to enhance the prediction accuracy while keeping a reasonable computational complexity.

Coverage optimization is a crucial task for a radio network operator. An accurate coverage estimation is a key prerequisite for efficient coverage analysis and optimization. In this paper, we propose a coverage prediction method based on statistical modeling of the wireless environment. We build a Radio Environment Map by interpolating geo-located measurements using the Kriging spatial prediction technique. Moreover, as we perform Kriging on massive observation datasets obtained through field measurement campaigns, we use Fixed Rank Kriging, to reduce the complexity of the Kriging algorithm. We apply the FRK algorithm for Long Term Evolution (LTE) network coverage prediction. We consider as observation data, the coverage measurements obtained by operational drive tests in a rural area. Numerical results show that by using the FRK algorithm, we fulfill a good trade-off between computational complexity and prediction accuracy.

During the last decade a lot of effort has been spent on cellular network optimization to improve network capacity and end-user Quality of Service (QoS). Coverage analysis remains as one of the essential topics on which mobile operators still need innovation in terms of performance and cost. Manual coverage analysis is an inefficient and costly task. Radio Environment Maps (REMs) is an efficient coverage analysis solution for present-day cellular networks. REM concept consists of spatially interpolating geo-located measurements to build the whole coverage map using a spatial interpolation technique originating from geo-statistics. Kriging is such a powerful technique which results in high performance in terms of prediction quality. However, this method is costly in terms of computational complexity especially for large datasets: computational complexity of Kriging is O(n3) where n is the number of measurements. This paper proposes the application of a variant of Kriging, Fixed Rank Kriging (FRK), to coverage analysis in order to reduce the computational complexity of the spatial interpolation while keeping an acceptable prediction error.

In this paper, a distributed stochastic approximation algorithm is studied. Applications of such algorithms include decentralized estimation, optimization, control or computing. The algorithm consists in two steps: a local step, where each node in a network updates a local estimate using a stochastic approximation algorithm with decreasing step size, and a gossip step, where a node computes a local weighted average between its estimates and those of its neighbors. Convergence of the estimates toward a consensus is established under weak assumptions. The approach relies on two main ingredients: the existence of a Lyapunov function for the mean field in the agreement subspace, and a contraction property of the random matrices of weights in the subspace orthogonal to the agreement subspace. A second-order analysis of the algorithm is also performed under the form of a central limit Theorem. The Polyak-averaged version of the algorithm is also considered.

The paper addresses the convergence of a decentralized Robbins-Monro algorithm for networks of agents. This algorithm combines local stochastic approximation steps for finding the root of an objective function, and a gossip step for consensus seeking between agents. We provide verifiable sufficient conditions on the stochastic approximation procedure and on the network so that the decentralized Robbins-Monro algorithm converges to a consensus. We also prove that the limit points of the algorithm correspond to the roots of the objective function. We apply our results to Maximum Likelihood estimation in sensor networks.

Issues involving missing data are typical settings where exact inference is not tractable as soon as nontrivial interactions occur between the missing variables. Approximations are required, and most of them are based either on simulation methods or on deterministic variational methods. While variational methods provide fast and reasonable approximate estimates in many scenarios, simulation methods offer more consideration of important theoretical issues such as accuracy of the approximation and convergence of the algorithms but at a much higher computational cost. In this work, we propose a new class of algorithms that combine the main features and advantages of both simulation and deterministic methods and consider applications to inference in hidden Markov random fields (HMRFs). These algorithms can be viewed as stochastic perturbations of variational expectation maximization (VEM) algorithms, which are not tractable for HMRF. We focus more specifically on one of these perturbations and we prove their (almost sure) convergence to the same limit set as the limit set of VEM. In addition, experiments on synthetic and real-world images show that the algorithm performance is very close and sometimes better than that of other existing simulation-based and variational EM-like algorithms.

MOTIVATION: One important aspect of data-mining of microarray data is to discover the molecular variation among cancers. In microarray studies, the number n of samples is relatively small compared to the number p of genes per sample (usually in thousands). It is known that standard statistical methods in classification are efficient (i.e. in the present case, yield successful classifiers) particularly when n is (far) larger than p. This naturally calls for the use of a dimension reduction procedure together with the classification one. RESULTS: In this paper, the question of classification in such a high-dimensional setting is addressed. We view the classification problem as a regression one with few observations and many predictor variables. We propose a new method combining partial least squares (PLS) and Ridge penalized logistic regression. We review the existing methods based on PLS and/or penalized likelihood techniques, outline their interest in some cases and theoretically explain their sometimes poor behavior. Our procedure is compared with these other classifiers. The predictive performance of the resulting classification rule is illustrated on three data sets: Leukemia, Colon and Prostate.