This webpage provides information for the 20th Heidelberg Summer School on the topic

AI in Astronomy

The school took place September 8 - 12, 2025 at the "Mathematikon", INF 205, 69120 Heidelberg, Germany.

Scientific organizing committee:
Tobias Buck (IWR/ITA)

Aleksandra Ciprijanovic (Fermilab)

Ioana Ciuca (KIPAC

Carolina Cuesta-Lazaro (Harvard)

Matthew Ho (Columbia University)

Francois Lanusse (CNRS)

Challenges in Modern Astronomy
Astronomy today faces unprecedented challenges as observational datasets grow exponentially in size and complexity. Extracting meaningful insights from massive surveys, simulating intricate cosmic processes, and addressing uncertainties in astrophysical models are some of the key difficulties. Understanding gravitational waves, modeling galaxy evolution, and analyzing exoplanet atmospheres require sophisticated computational techniques beyond traditional methods.

Advancements in AI and Machine Learning
Recent breakthroughs in artificial intelligence and machine learning have opened new frontiers in tackling these astronomical challenges. AI-driven techniques enable efficient data processing, enhance simulations, and improve predictive modeling. Machine learning algorithms, particularly deep learning and probabilistic inference methods, allow astronomers to analyze high-dimensional datasets, generate realistic cosmic models, and infer hidden parameters with greater accuracy.

Impact of AI on Astronomy
By leveraging AI advancements, astronomers can now extract deeper insights from observational data, model complex astrophysical systems with improved precision, and enhance scientific discovery. AI applications range from detecting exoplanets and classifying galaxies to modeling black hole mergers and reconstructing cosmological structures. These methodologies not only refine our understanding of the universe but also pave the way for future discoveries.

Equipping Students with Interdisciplinary Skills
The IMPRS Summer School 2025 is designed to provide participants with hands-on experience in modern AI techniques tailored for astrophysical research. Through lectures, interactive tutorials, and collaborative projects, students will gain expertise in state-of-the-art machine learning frameworks, simulation methodologies, and uncertainty quantification approaches. This interdisciplinary training will prepare attendees to integrate AI into their research and contribute to the advancement of astronomical sciences.

Topics Covered
The summer school will feature a series of lectures and hands-on sessions led by leading experts in AI and astrophysics. The key topics include:

Differential programming and generative models:
normalizing flows and diffusion models and their applications in stellar spectra analysis and astrophysical modeling, among others

Advanced probabilistic approaches for astrophysical data interpretation:
among others Simulation-Based Inference and its application to galaxies, the interstellar medium, Cosmology, and large-scale structure

Representation learning, multimodal learning, Foundation models and large language models in astronomy:
exploring applications of pre-trained AI models in astronomical research

Robustness of deep learning models, domain adaptation and interpretable machine learning models for physics discovery:
among others gravitational lens parameter inference or cosmological parameter estimation with Graph Neural Networks

  Lecture slides / material (pw protected)

  School gallery (pw protected)

  List of participants

  Summer school poster:

          Size:     A2 jpg,     A3 pdf,     A2 pdf,