What: Astrophysics Colloquium

When: May 18th, 2026, 16:00-17:00 CET

Where: Building C - Room 248 (first floor)

Speaker: Prof. Miguel Á. Sánchez-Conde, Universidad Autónoma de Madrid

Title: Searching for dark matter in a clumpy universe

Abstract: Dark matter halos in the standard LCDM model contain abundant substructure, or subhalos, which are crucial for indirect dark matter searches through annihilation signals. The most massive subhalos host dwarf galaxies, currently prime targets, while smaller dark subhalos may also provide excellent opportunities due to their proximity and high dark matter content. Moreover, the clumpy distribution of subhalos enhances the overall annihilation signal. This talk will present numerical work on Milky Way–like subhalo populations and discuss their impact on gamma-ray dark matter search strategies with instruments such as Fermi LAT, MAGIC, and CTAO, including recent searches and derived constraints.

What: Astrophysics Colloquium

When: Mar 30th, 2026, 16:00-17:00 CET

Where: Building C - Room 131 (ground floor)

Speaker: Dr. Michele Ginolfi, University of Florence

Title: Decoding Galaxy Properties with Machine Learning and Simulation-Based Inference

Abstract: The era of large-scale astronomical surveys demands innovative approaches for the rapid and accurate analysis of vast spectral datasets. A promising direction to address this challenge lies in the synergy between physics-based simulations and machine learning. Astronomy, like many other scientific disciplines, relies heavily on simulations to encode our physical understanding of complex processes, generate hypotheses, design experiments, and explore causal relationships. Increasingly, this knowledge is embedded in forward models that produce realistic synthetic spectra, images, and data-cubes. A central and long-standing challenge, however, is the solution of the "inverse problem": inferring the underlying physical parameters from observational data. In this talk, I will present a powerful framework to tackle this problem based on simulation-based inference (SBI) accelerated by invertible neural networks. Thanks to their amortised nature, these methods offer significant advantages over traditional Bayesian approaches: once trained, they enable fast inference and naturally yield complex, multi-modal posterior distributions conditioned on observables. I will show  preliminary results based on simulated spectra for upcoming instruments, along with ongoing SBI applications aimed at probing the interstellar medium in galaxies. Finally, I will discuss key open challenges, including model interpretability and domain adaptation, with particular emphasis on bridging the gap between synthetic training data and real observations.

What: phenomgrav astrophysics seminar

When: Mar 30th, 2026, 14:00-15:00 CET

Where: Building C - Room 248  (first floor)

Speaker: Dr. Josiel Mendonça Soares de Souza -  Universidade Federal do Rio de Janeiro (Brazil)

Title: Beyond the Fisher Approximation in Gravitational Wave Data Analysis

Abstract: In the next decade, new gravitational-wave detectors (like Einstein Telescope and Cosmic Explorer) will operate with a sensitivity gain of about one order of magnitude compared to current instruments. This improvement will enable the detection of binary black hole coalescences up to redshift z∼10, as well as hundreds of merger events per month. Such capabilities will make these detectors powerful instruments for cosmology. The main challenge lies in data analysis: Bayesian inference of binary black hole signals requires exploring a 15-dimensional parameter space with highly complex waveform models. One possible way to address this problem is to rely on approximate methods, such as the Fisher approach, which assumes that gravitational-wave posteriors are Gaussian. However, this approximation fails when the posteriors exhibit multimodality or strong non-Gaussian features. This limitation can be alleviated by including higher-order terms in the Taylor expansion of the log-likelihood. In this work, we demonstrate the effectiveness of the DALI (Derivative Approximation for Likelihoods) approach, which provides reliable gravitational-wave posteriors while remaining computationally efficient.