Data Assimilation and Machine Learning

The goal of this study group is to deepen understanding of data assimilation (DA) and machine learning (ML), and to explore their integration for diverse real-world applications.

Objectives

The primary objective of this study group is for researchers in the fields of data assimilation (DA) and machine learning (ML) to deepen their understanding and knowledge in these areas and to promote the discussions toward integration of DA and ML for their broad applications.

DA leverages physical models of a system’s dynamics to estimate its internal state, which may not be directly observed. In environments with limited observational data, DA is essential for inferring the current state and making accurate predictions. When ample observational data is available, a purely data-driven approach, such as machine learning, becomes more viable. This shift has been demonstrated in global weather prediction applications.

In this study group, we aim to explore the relationship and potential integration between these two methodologies across various problem domains.

Facilitators:: Shungo Tonoyama (RIKEN iTHEMS) – Contact: shungo.tonoyama@riken.jp; Yuta Tarumi (RIKEN iTHEMS); Ruiming Li (RIKEN iTHEMS); Kenji Okubo (RIKEN iTHEMS)