3IA PhD/Postdoc Seminar #54

Flash presentation | Unveiling Hidden Structures in the Main Belt: A Probabilistic Framework for Asteroid Families

Abstract: The identification of asteroid families is a key question in planetary sciences offering crucial insights into the collisional and dynamical history of the asteroid Main Belt (MB). These families, originating from the fragmentation of parent bodies due to catastrophic collisions, form dense clusters in orbital proper elements space. Over time, the non-gravitational Yarkovsky effect induces a semi-major axis drift, producing the characteristic V-shaped patterns in the (semi-major axis, absolute magnitude) plane.
In this study, we propose a new probabilistic approach for identifying asteroid families in the MB,
using model-based clustering. We model the observed population of the MB as a mixture of skewed-t
distributions coupled with a gaussian distribution which captures the Yarkovsky-driven semi-major axis evolution. This model also includes a uniform background component for the primordial asteroid population. The parameters are estimated using the Expectation-Maximization (EM) algorithm.

Frugal Deep Learning for Shockable Rhythm Detection in Ultra-Portable Defibrillators

Abstract: In collaboration with the 3IA Techpool team and the startup Inn'Pulse, this seminar will present our work on frugal artificial intelligence for real-time detection of shockable cardiac rhythms, designed for integration into next-generation ultra-portable automated external defibrillators (AEDs). By optimizing deep neural networks with neural architecture search and quantization-aware deployment, we demonstrate that lightweight ECG classifiers can meet international AED performance standards while running on low-power microcontrollers. The presentation will cover the full pipeline, from ECG processing and model training to robustness evaluation, interpretability analysis, and embedded deployment with energy measurements. We highlight trade-offs between speed, accuracy, and efficiency under design and hardware constraints.

Studies Show That Whatever I Say is True:  Source-Based Arguments and Citation Misuse

Abstract: Citations are widely regarded as markers of credibility, yet they are frequently misused in both academic discourse and public debate. This talk examines how and why such misuse occurs. Drawing on insights from argumentation theory, I will outline common ways in which valid data can be used to support flawed or misleading conclusions. Building on this framework, I explore how these theoretical tools can inform more effective fact-checking practices, particularly in the context of scientific misinformation. I will also discuss the limitations of current technological approaches and consider additional strategies for reducing the misuse of citations in contemporary information environments.

Open discussion about all the contributions