EAS 2022

Organiser

Host

Partner

Special Session SS24 28 June 2022

Critical Challenges for Machine Learning in Astronomy

Aims and scope

The success of machine learning algorithms in astronomy is undeniable. Deep learning has boosted the accuracy, speed and scale of data pipelines and analysis tasks across disciplines, from solar physics to cosmology. However, critical challenges still remain.

In this special session the best experts worldwide will discuss key challenges and solutions of data driven approaches applied to astronomy. By bringing together an ensemble of researchers developing cutting-edge techniques, we will improve the European community's awareness of the practical challenges in machine learning and we will meaningfully discuss and address their possible solutions. We will share best practices and give young researchers the opportunity to present new ideas and build future collaborations.

Programme

The program will consist of two 1.5h blocks, including invited talks, contributed talks, and poster presentations. The session will conclude with a 30 min open debate lead by expert panelists.

We will discuss recent work on these challenges, focusing (but not exclusively) on four main topics:

Uncertainty Estimation - an essential aspect of research which is still under-developed and even ignored
Data Generalization - e.g. domain adaptation from simulated images or different surveys, unbalanced data
Learning without assuming any implicit data distribution, e.g. using likelihood-free inference or generative adversarial neural networks
Identifying features in low-contrast images or noisy data