Data science and machine learning for climate research

While we are collecting and storing more and more data, understanding and interpreting the underlying processes and relationships between different observations remains a challenging problem.

The Future Earth Research School on “Data Science and Machine Learning for Climate Research” aims to provide a comprehensive overview of the mathematical foundations of machine learning, with a particular focus on methods for the analysis of complex dynamical systems.

Topics of interest include model reduction, system identification, estimation of associated transfer operators such as the Koopman operator, control, uncertainty quantification, deep learning, but also applications in climate science and sustainability.

Bertinoro (FC)

Location

5th – 17th June 2023

Dates

88 h

Total number of hours

20

Number of participants

€ 400,00

Registration Fee

March 6th

Application deadline

While we are collecting and storing more and more data, understanding and interpreting the underlying processes and relationships between different observations remains a challenging problem. The Future Earth Research School on “Data Science and Machine Learning for Climate Research” aims to provide a comprehensive overview of the mathematical foundations of machine learning, with a particular focus on methods for the analysis of complex dynamical systems and time-series data. The course will cover the following topics:

  • Reduced order modeling and system identification,
  • Koopman operator theory,
  • kernels and Gaussian processes,
  • deep learning,
  • quantum computing for classical dynamical systems,
  • stochastic differential equations and Markov state models,
  • agent-based models,
  • applications in climate science and sustainability.

The Summer School is mainly geared towards PhD students and postdocs in Mathematics or Climate Science with a strong interest in data science and machine learning. One of the main objectives is to learn about different modeling frameworks and data-driven methods for the analysis of large data sets. Applications include model reduction (e.g., identifying dynamically relevant variables), forecasting, control, and change-point detection. The course requires basic programming experience in Python (e.g., numpy, scipy, matplotlib).

The course will include lectures given by leading researchers in the fields of machine learning, dynamical systems theory, statistics, and climate science as well as seminars, hands-on sessions, and group projects. The lectures will take place in the morning, while the afternoons are reserved for tutorials, programming exercises, discussions, and group work.

The goal of the course is to first learn about state-of-the-art machine learning methods, to implement these methods in Python, and to then apply them to real-world problems. Furthermore, we will discuss novel applications of machine learning approaches, but also limitations and open problems.

Stefan Klus – Director of the Course

Stefan Klus is an associate professor in mathematics at the Heriot-Watt University, Edinburgh. He has several years of experience developing data-driven methods for the analysis of high-dimensional dynamical systems, with applications in molecular dynamics, fluid dynamics, quantum mechanics, agent-based modeling, and climate science. Before he joined the Heriot-Watt University, Dr. Klus worked at the University of Surrey, the Freie Universität Berlin, and the United Technologies Research Center, where he provided expertise in the areas of data science, machine learning, parallel computing, spectral and algebraic graph theory, and control of multi-agent systems, focusing on the analysis and optimization of numerical algorithms.

Houman Owhadi

Houman is professor of applied and computational mathematics and control and dynamical systems at the California Institute of Technology. His interests include uncertainty quantification, numerical approximation, statistical inference/learning, data assimilation, stochastic and multiscale analysis. His research is focused on solving numerical approximation problems as learning problems, learning problems as numerical approximation problems, and uncertainty quantification problems as adversarial games. He was a plenary speaker at SIAM CSE 2015, a tutorial speaker at SIAM UQ 2016, the recipient of the 2019 Germund Dahlquist Prize (SIAM), and a SIAM Fellow (class of 2022). His research is supported by DARPA, the Department of Energy, NASA/JPL, AFOSR, ONR, National Nuclear Security Administration, Los Alamos National Laboratory, Beyond Limits, and the National Science Foundation.

Joanna Slawinska

Joanna Slawinska is a Research Assistant Professor in the Department of Mathematics at Dartmouth College. She has a PhD in computational geophysics, in collaboration with the National Center for Atmospheric Research. She has obtained postdoctoral research training in applied mathematics from Courant Institute of Mathematical Sciences at New York University and has recently worked also as a Senior Researcher at the Department of Computer Sciences at the University of Helsinki. Joanna Slawinska has extensive expertise on operator theoretic approaches  for dynamical systems, as well as extraction and predictions of spatiotemporal patterns in turbulent flows. Her interests include also predictability of high-dimensional and time-resolved datasets, and more recently quantum computing.

Recently published work concerns nonparameteric predictions of tropical dynamics, data-driven schemes for approximating Koopman operators for climate dynamics, and novel frameworks for modeling nonlinear dynamics on quantum computers.

Feliks Nüske

Feliks Nüske received his Ph.D. in applied mathematics from Freie Universität Berlin in 2017. He then joined the Center for Theoretical Biological Physics at Rice University, U.S., for a postdoc. After a second postdoc in the Department of Mathematics at Paderborn University, Germany, he joined the Max-Planck-Institute for Dynamics of Complex Technical Systems, Magdeburg, Germany, as a research group leader. His research is on data-driven methods for modelling and simulation of molecular systems.

Location

The course will be held in The University Residential Center of Bertinoro

Bertinoro is situated halfway between the cities of Forlì and Cesena, 6 km away from SS9 (Via Emilia). Forlì is the town of reference for transport by train and bus to and from Bertinoro.

Accomodation

The accomodation will be at the University Residential Center of Bertinoro. Please remember that participation in presence is mandatory.

The school is offering a limited financial assistance to cover the accomodation costs at the Center of Bertinoro over the entirety of the two weeks stay. If you are interested, please check the section below.

For those who won’t get the financial assistance, all the costs will be comunicated during the selection process.

Lunches and dinners will be provided and offered by the school for all the students. Students are free to organize themselves at their own expense upon notice. The school will not cover any extra costs.

Transport

Nearest airport: Bologna’s airport “Guglielmo Marconi”

Nearest train station: Forlì Station (20 min. far from Bertinoro by car)

On how to reach the Centre, please consult this link

Given that most of the students will arrive in Bologna – especially from abroad – the school will organize a shuttle who will bring students from Bologna to Bertinoro. More details will be given to the selected candidate.

 

 

The school offer a limited financial assistance, based on the applicant’s needs. It will cover accommodation expenses in Bertinoro. If interested, please specify it in the application form in the dedicated section. You can also leave a statement (max 200 words) explaining why you should benefit from this assistance.

The assistance WILL NOT cover travel costs.

Since there is a limited number of places, you are encouraged to apply early to avoid disappointment.

If accepted, you will receive an email confirming your financial assistance by secretariat@fersschool.it

Start typing and press Enter to search