Speaker : David Cornu (Obs de Paris)
Title: Winning the SKA Science Data Challenge 2 with a fast Deep Learning object detector
Abstract:
With its 1 TB simulated data cube of HI line emission, the SKA Science Data Challenge 2 (SDC2) is getting closer to the difficulty of real upcoming SKA observation analysis. Even if the type of task to perform in the SKA SDCs are rather classical (detection, classification, parameter extraction, etc.) modern datasets have become heavily demanding for classical approaches due to size and dimensionality. It is not a surprise then, that many astronomers started to focus their work on Machine Learning approaches that demonstrated their efficiency in similar applications. However, hyperspectral images from astronomical interferometers are in fact very different from images used to train state-of-the-art pattern recognition algorithms, especially in terms of noise level, contrast, object size, class imbalance, spectral dimensionality, etc. As a direct consequence, these methods do not perform as well as expected when directly applied to astronomical datasets. In this context, the MINERVA (MachINe lEarning for Radioastronomy at obserVatoire de PAris) project has assembled a team to participate in the SDC2 with the objective of developing innovative Machine Learning methods that better suit the needs of astronomical images.
In this presentation, I will describe the work we have made on implementing a modern YOLO (You Only Look Once) CNN object detector inside our custom framework CIANNA (Convolutional Interactive Artificial Neural Networks by/for Astrophysicists) and describe the modifications and tuning that allowed us to reach the first place of the SKA SDC2. I will start by discussing the strengths and weaknesses of this type of method in comparison to more widely adopted Region-Based CNN detectors (Faster R-CNN, Mask R-CNN, …). I will also review the motivation and the effect of the numerous changes we made on the method (data quantization, 3D convolution, layer architecture, detection layout to manage blending, objectness decomposition, IoU selection, additional parameter inference, …) in order to apply it to both SDC1 and SDC2, and identify what are the present limits as well as some tracks for further improvements. I will detail the computational efficiency of the method (with GPU acceleration) and discuss its scaling capabilities for upcoming challenges or datasets. Finally, we will comment on how this methodology could be used to analyze the actual data from SKA pathfinders or any other similar astronomical dataset and how it could be used to merge knowledge and information from multiple datasets at the same time.
Speaker : Martin Turbet
Title: Le modèle « générique » de climats planétaires et panorama de ses applications
Abstract: Le modèle générique de climats planétaires ou « Generic PCM » est un code communautaire, développé principalement et historiquement au LMD, et dont l’objectif est de simuler l’ensemble des processus physiques et chimiques opérant dans les atmosphères planétaires. Ses applications sont nombreuses : étude de la dynamique atmosphérique de Jupiter, Saturne et des géantes glacées, formation de brumes photochimiques sur Titan, évolution couplée de l’atmosphère et des glaces sur Pluton et Triton, paléoclimats de Mars, la Terre et Vénus, climats et observabilité des exoplanètes (des plus froides et petites aux plus grandes et chaudes), etc.
Après un bref aperçu du modèle et des briques qui le constituent, je vous présenterai un panorama de ses applications, avec une attention toute particulière sur les planètes et exoplanètes telluriques.
Speaker : Louis Amard
Title: The evolution of young low-mass stars : focus on rotation and activity
Abstract: Between its formation stage as an active accreting seed and today, the Sun underwent large structural changes as well as variation in magnetic activity, rotation rate and its relation to the surrounding environment. I will go through the different processes that are responsible for these changes and present our latest results on the subject. We will go from the early interaction between the star and its proto-planetary disc bathing in UV radiations emitted from the massive neighbours, to the internal mixing happening in the inner layers of solar-like stars and probed by asteroseismology. Finally, I will try to review the possible applications of this work to other types of stars at different stages of the evolution in the context of the current or future surveys.
Speaker : Olivia Venot (LISA)
Title: On the importance of chemical data for warm exoplanet atmospheres
Abstract: The very first data from JWST are finally arriving and show us the full potential of this telescope for the characterisation of exoplanet atmospheres. The interpretation of these data relies on atmospheric models, so it is of paramount importance that these models are reliable and robust. In this context I will show you the methodology applied to develop kinetic models adapted to the atmospheres of hot exoplanets, with physico-chemical data at high temperatures. I will also talk about the Hot Jupiter WASP-43b and the insights that JWST can give us into its atmosphere.
Speaker : Laurent Chemin (Universidad Andres Bello, Instituto de Astrofisica)
Title: Dynamics of the Milky Way and LMC as seen by Gaia
Abstract:
Speaker : Philippe Bonneton (EPOC, CNRS/U. Bordeaux)
Title: Tidal bore dynamics in estuaries
Abstract: A tidal bore is a sudden elevation of the water surface that travels upstream an estuary with the incoming flood tide. The formation and dynamics of this spectacular and fascinating phenomenon bring into play nonlinear wave interactions over a large range of spatiotemporal scales. A tidal bore is a small-scale estuarine process with scales of variation of order few seconds in time and dozen meters in space. This small-scale process, which generally occurs in the upper estuary, results from the nonlinear transformation of the tidal wave over long distance (several dozen of kilometers) and long period of time (several hours). The general bore phenomenon has been extensively studied in fluid mechanics and mathematics. On the other hand, the large-scale mechanisms which control tidal bore formation has so far received little attention. In this presentation we will analyze tidal bore formation from a unique long term database acquired during 4 campaigns in the two main French tidal-bore estuaries: the Seine and Gironde/Garonne estuaries. We will show that the tidal bore intensity is mainly governed by the dimensionless tidal range, which characterizes the local tidal wave nonlinearity. We will also show that the undular bore structure can be explained by two different wave processes depending on the bore intensity.
Speaker : Camille Bergez-Casalou (LESIA)
Title: Simultaneous accretion of giant planets or how giant planets fight for food
Abstract: Among the thousands of exoplanetary systems observed nowadays, some of them host multiple giant planets, like in our Solar System. These giants are probably formed from the same protoplanetary disc. I’m going to present our recent paper where we investigated, with the help of 2D hydrodynamical simulations, how the gas present in the protoplanetary disc is distributed and accreted by two planetary cores. We start with 20 Me cores undergoing runaway gas accretion and follow the evolution of the planets mass ratio. We find that when the planets accrete from the same disc, they end up with very similar masses, independently on the disc viscosity or the delay in accretion between the planets. We finish by suggesting different formation scenarios in order to explain the observed exoplanets distribution.
Speaker : Damien Gratadour
Title: Bridging Astronomy, HPC and AI in giant telescopes from the optical to radio frequencies
Abstract: Frontier computing has become a key enabler for scientific discovery, at the core of the next generation of extremely large multi-science hubs, such as the SKA Observatory or the European Extremely Large Telescope. I will present current efforts and the future roadmap we are contemplating, bridging astronomy with HPC and AI technologies to contribute to these major scientific infrastructures, in particular through the soon to be established ECLAT joint lab, a collaborative effort between CNRS, Inria and Atos. Beyond technical aspects related to increasing the scientific return of these facilities, I will also cover transverse topics such as energy efficiency and operational cost containment.
Speaker : Christine Ducourante (LAB)
Title: Gaia explore l’Univers profond: Galaxies, quasars et lentilles gravitationnelles
Abstract:
Speaker : Lars Bonne (USRA)
Title: Probing the assembly and dispersal of molecular clouds with the [CII] fine structure line
Abstract:
I will present high spectral resolution observations of the ionized carbon fine structure line at 158 μm ([CII]) from the SOFIA FEEDBACK legacy survey. [CII] is a unique tracer of the CO-dark molecular gas and photodissociation regions (PDRs) at the surface of molecular clouds. The FEEDBACK survey maps the [CII] line in 11 Galactic high-mass star forming regions. First, I will present the [CII] observations of the DR21 ridge in Cygnus-X. There, [CII] unveils that the assembly of this massive star forming ridge is the result of a high-velocity (~20 km/s) colliding flow. This collision also curves the magnetic field around the ridge and thus sets the cloud geometry. Further investigating the molecular cloud dynamics with IRAM 30m observations shows that this rapid mass accumulation initiates gravitational collapse of the full molecular cloud. Then, I will present studies focusing on regions that experience the effect of stellar feedback from high-mass stars. In these regions, [CII] unveils high-velocity mass ejection from the clouds that is not seen in CO. Quantifying the mass ejection rate shows that stellar feedback can disperse these star forming clouds in a few Myr. These results point to a consistent scenario for molecular cloud evolution and star formation on relatively short timescales. Far-infrared fine structure lines are thus important tools to pin down the physical processes that drive the evolution of the interstellar medium (ISM).