Workshops

W1: Recent developments in multiblock methods and the use of the RGCCA package, by Pr. Arthur Tenenhaus

You can submit workshop proposals until April 17, 2026.

Invited speakers

Heterogeneous data integration and fusion

Pr. Arthur Tenenhaus. Arthur Tenenhaus is professor of statistics/machine learning at CentraleSupelec (top engineering school in France) and member of the  ‘Signals and Statistics’ group of the Laboratoire des Signaux and Systèmes. He is also researcher affiliated to the Paris Brain Institute (ICM) at La Pitié-Salpêtrière Hospital and co-holder of the chair  APHP-CentraleSupelec- INRIA. His main research interest concerns the development of statistical framework for  the joint analysis of heterogeneous and complex data. This framework is usually motivated by applications  in molecular biology and neuroimaging.

Title: Structural equation modeling with factors and composites within the framework of the basic design

Summary: We study structural equation modeling with factors and composites in the framework of the basic design.  We propose two estimation methods. The first one is a new non-iterative approach based on singular  value decomposition calculations (SVD-SEM). This new approach produces consistent and asymptotically  normal estimators of the parameters. Then, we describe the restricted maximum likelihood approach in  the framework of the basic design. We evaluate the efficiency and usefulness  of SVD-SEM and ML-SEM  in a Monte Carlo simulation and on a case study.

Modeling and digital twins

Dr. David Camilo Corrales. David Camilo Corrales is a computer scientist with a PhD in Computer Science specialized in Artificial Intelligence from Universidad Carlos III de Madrid (Spain). He has completed three postdoctoral fellowships at INRAE Occitanie-Toulouse, where he developed machine learning models for pest, disease, and weed prediction in agronomic applications. He holds a bachelor’s degree in software engineering and a master’s degree in telematics engineering from Universidad del Cauca (Colombia). His applied research in AI spans different domains, including water quality, volcanology, and agronomy. Since 2021, he has focused on digital and cyber-physical systems in biotechnology. He has authored more than fifty publications in peer-reviewed journals and international conferences. Currently, he works at INRAE - Toulouse Biotechnology Institute (TBI), where he contributes to the development of digital twins, virtual sensors (soft sensors), and hybrid models integrating machine learning with dynamic process modeling. His expertise in MLOps pipelines for real-time deployment of AI-based solutions supports the creation of scalable, robust, and data-driven digital systems for complex cyber-physical infrastructures.

Title: Digital Twins and Industry 4.0: From Physical Models to Operational Intelligence with AI and MLOps

Abstract: The convergence of digital technologies with industrial biotechnology is transforming how modern bio-processes are designed, optimized, and operated. As the sector moves toward the vision of Industry 4.0, new approaches are required to accelerate research and development, reduce time-to-market, and enable autonomous, adaptive, and sustainable production systems. This keynote will explore how digital twins, AI-driven operational intelligence, and MLOps pipelines are reshaping bio-process engineering from early-stage design to real-time control. Drawing on insights from the European initiative BIOINDUSTRY 4.0, the presentation will show how data centric and hybrid modeling strategies combine physical models with machine learning to create scalable digital infrastructures. We will discuss how advanced data standards and interoperable metadata frameworks contribute to high-quality multi-scale datasets, the foundation for digital twins. These digital twins, together with virtual sensors and hybrid models deployed in real time, enable predictive control and accelerate optimization in bioprocess development.

Behavioral & physiological analysis in context

Dr. Catherine Soladié received her PhD in 2013 and is currently a researcher at CentraleSupélec and IETR. Her work focuses on the analysis of human behavior and physiological signals in real-world contexts. Her research lies at the intersection of signal processing, multimodal data analysis, and human-centered computing. She is particularly interested in understanding human emotions and how behavioral patterns and physiological responses can be jointly modeled to provide deeper insights into human activity, interaction, and health-related processes.

Title: Weakly Supervised and Personalized Emotion Modeling: From Subjective Perception to Individual Dynamics

Abstract: Understanding human emotions through behavioral and physiological signals is becoming increasingly important in domains such as health monitoring, affective computing, and adaptive human–machine interaction. Yet, capturing emotions in an objective and reliable way remains a fundamental challenge. Emotional states are deeply subjective, shaped by personal experience and individual perception, and they are therefore difficult to define, measure, and annotate consistently. In this context, unsupervised and weakly supervised learning approaches provide promising alternatives. By limiting the dependence on explicit emotion labels—which are often costly, noisy, or biased—these methods can help uncover meaningful patterns directly from multimodal data. A second major challenge lies in the strong variability between individuals. Emotional responses are highly personal, and absolute levels are often less informative than changes within a single person over time. This makes personalization essential, with a focus on capturing individual dynamics rather than applying one-size-fits-all models. By combining wearable sensors, digital assessment tools, and multimodal machine learning, this work aims to support more adaptive and robust emotion analysis in real-world settings.

Visual data processing & automated image-based analysis

Dr Rémy Sun is an ISFP research scientist in the MAASAI team at Inria d’Université Côte-d’Azur since October 2024. He previously completed his PhD with Matthieu Cord and Nicolas Thome at Sorbonne University's MLIA team on content combination strategies for image classification, and worked on deep learning research in European laboratories such IST Austria and the Max Planck Institute for Intelligent Systems. His current research interests focus mostly on knowledge injection in neural networks, multimedia interpretation, deep learning for physics and conditional/guided diffusion models.

Title: From Computer Vision to Marine Biodiversity Monitoring.

Abstract: The health of marine ecosystems is under unprecedented threat from human-induced pressures such as habitat degradation overfishing and climate change. However, traditional monitoring methods are labour-intensive, limited by diver availability, and subject to observer bias. This presentation will discuss how deep learning tools from Computer Vision can be adapted to automate this monitoring pipeline through the use of object detectors to find fishes in diver videos, automated techniques to count fishes seen in such videos and provide insight into the effects of Marine Protected Areas on the French Rivieira. These efforts establish a robust, scalable framework for ecological monitoring that minimizes human bias, enhances data reproducibility, and enables high-resolution, long-term biodiversity assessments in a step towards data-driven conservation and sustainable management of vulnerable marine ecosystems.

Data quality, reliability and reproducibility

Pr. Anne-Laure Boulesteix. After a double-degree in engineering and mathematics from the Ecole Centrale Paris and the University of Stuttgart (2001), a PhD in statistics (2005) from the Ludwig Maximilian University (LMU) of Munich and a postdocs in medical statistics, Anne-Laure Boulesteix joined the Medical School of the University of Munich as a junior professor (2009) and professor (2012). She is working at the interface between biostatistics, machine learning and medicine with a particular focus on metascience. She is a steering group member of the STRATOS initiative, former president of the German Region of the International Biometric Society (2023-2025) and holder of a Reinhart-Koselleck grant (2025-2030) from the German Research Foundation (DFG). One of her main research interests is the methodology of reliable comparisons of statistical methods. , Department of Medical Informatics, Biometry and Epidemiology of the Faculty of Medicine, LMU Munich.

Title: Towards appropriate study designs and reliable empirical evidence in methodological (biostatistics) research.

Abstract: Statisticians are often keen to analyze the statistical aspects of the so-called “replication crisis in science“. They condemn fishing expeditions and are involved in designing studies across empirical scientific fields applying statistical methods, such as life sciences. But what about good practice issues and study designs in their own - methodological - research, i.e. research considering statistical (or more generally, computational) methods as research objects? When developing and evaluating new statistical methods and data analysis tools, do statisticians and data scientists adhere to the good practice principles they promote in fields which apply statistics and data science? In the last few years, statisticians have started to make substantial efforts to address what may be called the replication crisis in the context of methodological research in statistics. In this presentation, I will give an overview of recent positive developments including own projects and projects by others. Promising concepts include for example study protocols, confirmatory research, appropriate handling of method failure, a phases classification system for methodological studies, and real data based simulation designs.

Scientific program

The details of the scientific program will be communicated later. The agenda below is provisional.

Monday 5^th October

 Tuesday 6^th October

Wednedsay 7^th October