CONFERENCE
Adaptive and High-Dimensional Spatio-Temporal Methods for Forecasting
Méthodes spatio-temporelles adaptatives et en grande dimension pour la prédiction
26 – 30 September 2022
Scientific Committee
Comité scientifique
Richard Davis (Columbia University)
Anne-Laure Fougères (Université Lyon 1)
Marc Genton (King Abdullah University of Science and Technology)
Claudia Klüppelberg (Technical University of Munich)
Olivier Wintenberger (Sorbonne Université)
Organizing Committee
Comité d’organisation
Jean-Marc Bardet (Université Paris 1 – Panthéon Sorbonne)
Philippe Naveau (CNRS, LSCE IPSL)
Rainer von Sachs (Catholic University of Leuven)
Suhasini Subba Rao (Texas A&M University)
Almut Veraart (Imperial College London)
This conference proposal follows the CIRM conference ”New results on time series and their statistical applications” organized from September 14 to 18, 2020. During the latter, the focus was on time series, typically sequences of random variables, with an interest in problems related to non-stationarity and dependency, extremes, asymptotic behavior of estimators and test statistics. But it appeared to us, also through presentations outside this framework, that it would be extremely fruitful to extend the spectrum of the work presented to sequences of random vectors, as well as sequences of functions or random elements of spaces with certain geometries (typically manifolds such as the one of positive definite matrices). The field of applications of such models is multiple: let us quote for instance sequences of several dependent economic indicators or financial indices, electricity consumption in different countries and different scales, annual climate curves in various regions, trajectories of planes, evolution of the traffic in a road network, etc. The question of forecasting time-dependent mathematical objects is obviously a central issue. The ”classical” approach of mathematical statistics referring to the notion of model, parametric or not, and allowing forecasting from probabilistic type hypotheses can provide a benchmark that may encounter computational troubles to deal with high-dimensional models. However it appears that the recent developments for handling complex data in machine learning can be use to incorporate the contributions of a huge number of covariates (location, week of the day, etc.) and infinite dimensional informations (local expansions, derivatives, eigen space, etc.) in order to improve the forecast. The aim is then to design recursive and/or adaptive algorithms, taking into account or not multiple underlying models and able to handle complex data in an automatic way. We can thus think about predictor’s aggregations, online gradient based algorithms, etc.
Cette proposition de conférence fait suite à la conférence du CIRM ”Séries chronologiques : nouveaux résultats et applications statistiques” organisée du 14 au 18 septembre 2020. Au cours de cette dernière, l’accent a été mis sur les séries temporelles, généralement des suites de variables aléatoires, avec un intérêt pour les problèmes liés à la non-stationnarité et à la dépendance, aux extrêmes, au comportement asymptotique des estimateurs et statistiques de test. Mais il nous est apparu, également à travers des présentations en dehors de ce cadre, qu’il serait extrêmement fructueux d’étendre le spectre des travaux présentés à des suites de vecteurs aléatoires, ainsi qu’à des suites de fonctions ou d’éléments aléatoires d’espaces ayant certaines géométries (typiquement des variétés comme celle des matrices définies positives). Le champ d’application de tels modèles est extrêmement large : citons par exemple les évolutions conjointes de plusieurs indicateurs économiques ou financiers, la consommation d’électricité dans différents pays et à différentes échelles, les courbes climatiques annuelles dans différentes régions, les trajectoires des avions, l’évolution du trafic dans un réseau routier… La question de la prévision d’objets mathématiques dépendants du temps est évidemment centrale. L’approche ”classique” des statistiques mathématiques faisant référence à la notion de modèle, paramétrique ou non, et permettant de faire des prévisions à partir d’hypothèses de type probabiliste peut fournir une référence, qui peut cependant rencontrer des difficultés de calcul pour traiter les modèles de grande dimension. Les récents développements pour le traitement de données complexes dans l’apprentissage statistique peuvent être utilisés pour intégrer les contributions d’un nombre énorme de covariables (lieu, semaine du jour…) et d’informations de dimensions infinies (expansions locales, dérivées, sous-espaces propres…) afin d’améliorer la prévision. L’objectif est alors de concevoir des algorithmes récursifs et/ou adaptatifs, prenant en compte ou non de multiples modèles sous-jacents et capables de traiter des données complexes de manière automatique. On peut ainsi réfléchir à des agrégations de prédicteurs, des algorithmes basés sur des gradients en ligne…
SPEAKERS
Omer Bobrowski (Technion Israel Institute of Technology) Persistent homology, time series, and universality
David Bolin (King Abdullah University of Science and Technology) Gaussian Whittle-Matern fields on metric graphs
Alessia Caponera (EPFL) Functional Estimation of Anisotropic Covariance and Autocovariance Operators on the Sphere
Carsten Chong (Columbia University) Estimating roughness in volatility
Daniel Cirkovic (Texas A&M University) Likelihood-based Changepoint Detection in Preferential Attachment Networks
Alexandre Constantin (Université Grenoble Alpes, CNRS) Mixture of Multivariate Gaussian Processes for Classification of irregularly sampled time-Series: Application to Massive Satellite Image
Rainer Dahlhaus (University Heidelberg) Adaptive estimation for locally stationary processes
Anna van Delft (Columbia Univeristy) A general framework to quantify deviations from structural assumptions in the analysis of nonstationary function-valued processes
Ania Dudek (AGS University of Science and Technology in Krakow) Modeling ECG signals using OACS processes
Marie Düker (Cornell University) High-dimensional latent Gaussian count time series
Jean-Marc Freyermuth (I2M, Aix-Marseille Université) Some contributions to harmonizable time series analysis
Yannig Goude (Université Paris-Saclay) Hierarchical transfer learning with applications for electricity load forecasting
Christian Hirsch (Aarhus University) Topology-based goodness-of- t tests for sliced spatial data
Stephan Huckemann (Georg-August-University of Goettingen) PCA and Clustering on Manifolds
William Kengne (CY Cergy Paris Université) Statistical learning for psi-weakly dependent processes
Rebecca Killick (Lancaster University) Detecting changes in mixed-sampling rate data sequences
Claudia Kirch (Otto-von-Guericke University Magdeburg) Scan statistics for the detection of anomalies in random fields
Tobias Kley (Georg-August-University of Goettingen) Change point analysis with irregular signals: when did the COVID-19 pandemic start?
Jonas Krampe (University of Mannheim) Frequency Domain Statistical Inference for High-Dimensional Time Series
Jeongjin Lee (University of Namur) Partial Tail Correlation for Extremes
Dan Leonte (Imperial College London) Simulation and forecasting method for trawl processes and ambit fields
Anne Leucht (University of Bamberg) Mixing properties of non-stationary INGARCH(1,1) processes
Yuan Li (Imperial College London) Limit theorems for the realised semicovariances of multivariate Brownian semistationary processes
Tomas Masák (EPFL) Separable Expansions for Covariance Estimation via the Partial Inner Product
Linda Mhalla (EPFL) Extremal connectedness of hedge funds
Hernando Ombao (King Abdullah University of Science and Technology) Exploring General Dependence in a Brain Network
Thomas Opitz (INRAE Avignon) Spatiotemporal modeling of extreme events: a state of the art
Victor Panaretos (École Polytechnique Fédérale de Lausanne) The Extrapolation of Correlation
Olivier Pache (Université de Genève) Neural Networks for Extreme Quantile Regression with an Application to Forecasting of Flood Risk
Leo Pfitzner (EPFL) Temperature forecasting with expert aggregation
Anne Philippe (Université de Nantes) Detection of non-stationarity in dependent data
Daniel Rademacher (Heidelberg University) Inference for the Spectral Mean of Functional Time Series in Hilbert Space
Stefan Richter (University of Tennessee) Empirical process theory and oracle inequalities for locally stationary processes
Peter Robinson (London School of Economics and Political Science) Issues in Spatial Processes with Long Range Dependence
François Roueff (TelecomParisTech) Fractionally Integrated Autoregressive Moving Average processes valued in a separable Hilbert space
Leonardo Santoro (EPFL) Nonparametric Estimation for SDE with Sparsely Sampled Paths: an FDA Perspective
Orimar Sauri (Aalborg University) Nonparametric estimation of trawl processes: Theory and Applications
Johan Segers (UC Louvain) Modelling multivariate extreme value distributions via Markov trees
Kristin Strokorb (Cardiff University) Modelling multivariate extreme value distributions via Markov trees
Lionel Truquet (ENSAI) Asymptotics for estimating the Taylor’s law parameters in Ecology
Henrik Wiechers (Georg-August-University of Goettingen) Learning torus PCA based classification for multiscale RNA correction with application to SARS-CoV-2
Junho Yang (Academia Sinica) Baxter-type convergence results for a block Toeplitz system under long memory with an application to linear prediction problem
Yi Yu (University of Warwick) Change point inference in high-dimensional regression models under temporal dependence
SPONSORS
