7th edition of the Workshop on
Machine Learning and Data Assimilation for Dynamical Systems
MLDADS 2025
7-9 July 2025
7-9 July 2025
ICCS 2025 : https://www.iccs-meeting.org/iccs2025/
Previous editions of MLDADS:
Malaga, MLDADS 2024
Prague, MLDADS 2023
London, MLDADS 2022
Poland (online) MLDADS 2021
Amsterdam (online) MLDADS 2020
Faro, MLDADS 2019
Title: Machine Learning and Data Assimilation for Dynamical Systems - MLDADS 2025
Organisers:
Rossella Arcucci and the Data Learning working group at Imperial College London
Sibo Cheng, CEREA, ENPC, Instituit Polytechnique de Paris
Tobias Necker, European Centre for Medium-Range Weather Forecasts (ECMWF)
This workshop is supported by the Phi-Lab at ESA (European Space Agency)
Important Dates:
Paper submission: 31 January 2025 28 February 2025 14 March 2025
Notification to authors: 31 March 2025 4 April 2025
Camera-ready papers: 18 April 2025
Author registration: 31 March – 18 April 2025
Non-author registration (in-person only): 31 March – 2 June 2025
Conference sessions: 7-9 July 2025
Abstract: The field of dynamical systems modelling is advancing rapidly, yet significant challenges remain due to the inherent limitations of existing models—whether physics-based or data-driven—in capturing complex real-world phenomena with high accuracy. Data assimilation and data fusion have emerged as powerful methodologies for improving model accuracy through the integration of observational data. However, these techniques are often computationally intensive, posing limitations on scalability and real-time applicability. Concurrently, Machine Learning (ML) offers a suite of algorithms capable of enhancing task performance with the integration of larger, more diverse data sets. The growing convergence of ML with data assimilation and fusion promises transformative potential for modeling dynamical systems, driving both theoretical advances and practical applications.
The primary aim of this workshop is to convene researchers from data assimilation, machine learning and dynamical systems to bridge the gaps between these fields. By exploring how these complementary disciplines can accelerate research outcomes and impact, the workshop seeks to foster collaboration, share cutting-edge advancements, and tackle the computational challenges that have limited the application of data assimilation and fusion in high-dimensional complex systems as well as the application of machine learning for dynamical systems.
Identify the Shared Challenges: Discuss the primary computational and theoretical obstacles in data assimilation when applied to real-world dynamical systems and explore potential machine learning solutions.
Enhance Model Accuracy and Efficiency: Showcase how machine learning techniques—such as deep learning, reinforcement learning, and transfer learning—can improve the efficiency and scalability of data assimilation in dynamical system modelling.
Promote Interdisciplinary Collaboration: Encourage interdisciplinary partnerships to merge theory and practice, enabling novel approaches to modelling that can be deployed across diverse domains such as weather forecasting, fluid dynamics, climate science, and more.
Encourage Open-Source Solutions: Highlight open-source platforms and tools that integrate ML with data assimilation frameworks to democratise access to computational techniques and streamline innovation.
Call for Papers: The MLDADS workshop is to bring together contributions from the fields data assimilation, machine learning and dynamical systems to fill the gap between these theories in the following directions:
1) Machine Learning for Dynamical Systems: how to analyse dynamical systems on the basis of observed data rather than attempt to study them analytically; explore the role of generative models.
2) Data Assimilation for Machine Learning and/or Dynamical Systems: how well does the model under consideration (Machine Learning model and/or Dynamical System) represent the physical phenomena.
3) Machine Learning for Data Assimilation: how machine learning techniques—such as deep learning, reinforcement learning, and transfer learning—can improve the efficiency and scalability of data assimilation in dynamical system modelling.
4) Data Assimilation and Machine Learning for Dynamical Systems: how can tools from the interaction between the theories of Data Assimilation and Machine Learning be used to improve the accuracy of the prediction of dynamical systems.
Program Committee:
Celine Robardet, National Institute of Applied Science in Lyon, France.
Roland Potthast, Deutscher Wetterdienst (DWD), Germany.
Ionel Michael Navon, Florida State University, Florida US.
Marta Chinnici, National Agency for New Technologies, Energy and Sustainable Economic Development, Italy.
Andrew M. Moore, University of California Santa Cruz, California US.
Nancy Nichols, University of Reading, UK.
Livia, Marcellino, University of Naples Parthenope, Italy.
Tiziana Di Matteo, Kings College, UK.
Gianluca, Bontempi, Université Libre de Bruxelles, Belgium.
Massimiliano Pontil, UCL and Istituto Italiano di Tecnologia, UK and Italy.
Luca Magri, Cambridge University, UK.
Marco Gallieri, NNAISENSE, Switzerland.
Vishwas Hebbur Venkata Subba Rao, Argonne National Laboratory, Lemont, IL.
Gabriele Santin, Fondazione Bruno Kessler, Italy.
Alberto Carrassi, University of Bologna, Italy & University of Reading, UK.
During submission, you may select either a “Full/Short Paper” or a “Abstract Only” publication. You can also define your preference for an oral or poster presentation. Please understand that this is only a preference. The presentation type of accepted submissions will ultimately be defined by the Organizing Committee and communicated in the notification email.
While we encourage full paper submissions, the “Abstract Only” option caters to researchers who can only publish in specific journals or work for companies in circumstances such that they cannot publish at all, but still want to present their work and discuss it with their peers at ICCS. In the “Abstract Only” option, a short abstract is included in the conference program, but not in LNCS.
After the conference, the best papers will be invited for a special issue of the Journal of Computational Science (Impact Factor: 3.3)
The manuscripts of up to 15 pages, written in English and formatted according to the Springer LNCS templates, should be submitted electronically via Easychair.
The most recent versions of the Templates are available for download from this link.
The LaTeX2e Proceedings Templates are also available in the scientific authoring platform Overleaf.
Please also make sure to follow Springer’s authors’ guidelines, as well as Springer’s Book Authors’ Code of Conduct.
Papers must be based on unpublished original work and must be submitted to ICCS only. Submission implies the willingness of at least one of the authors to register and present the paper.
To promote open science and to facilitate reproducibility, ICCS encourages data and code sharing. Please keep this in mind when submitting your papers.
Deadlines for draft paper submission, notification of acceptance, camera-ready paper submission and registration may be found in the Important Dates section.
For the camera-ready version, please note that:
ICCS authors interested in open access for their papers (open choice) have the possibility to do so. When you submit your camera-ready paper, if you intend to make it open access, contact us and we will tell you how to proceed.
Springer encourages authors to include their ORCIDs in their papers.
Springer is now offering the inclusion of embedded videos in proceedings papers.
For information please contact Dr. Rossella Arcucci, r.arcucci@imperial.ac.uk