Dates: February, 3 - 7, 2020
Organisers: Martin Kilbinger, ...
Venue: Institut d'Astrophysique de Paris (IAP), 98bis bd Arago, 75014 Paris.
Local information: http://www.iap.fr/accueil/acces/acces.php?langue=en
Please add your name to the following google doc if you are planning to attend the meeting.
There is no registration fee. Coffee and snacks will be provided for the breaks. For lunch, participants are invited to go to the nearby restaurants, shops, or imbiss stands
(see http://www.iap.fr/vie_scientifique/colloques/Colloque_IAP/2018/i-practicalinfo.html#lunch for some ideas).
Galaxy cluster counts in bins of mass and redshift have been shown to be a competitive probe to test cosmological models. This method requires an efficient blind detection of clusters from surveys with a well-known selection function and robust mass estimates. The Euclid wide survey will cover 15000 deg2 of the sky in the optical and near-infrared bands, down to magnitude 24 in the H-band. The resulting data will make it possible to detect a large number of galaxy clusters spanning a wide-range of masses up to redshift ∼2. This paper presents the final results of the Euclid Cluster Finder Challenge (CFC). The objective of these challenges was to select the cluster detection algorithms that best meet the requirements of the Euclid mission. The final CFC included six independent detection algorithms, based on different techniques, such as photometric redshift tomography, optimal filtering, hierarchical approach, wavelet and friend-of-friends algorithms. These algorithms were blindly applied to a mock galaxy catalog with representative Euclid-like properties. The relative performance of the algorithms was assessed by matching the resulting detections to known clusters in the simulations. Several matching procedures were tested, thus making it possible to estimate the associated systematic effects on completeness to <3%. All the tested algorithms are very competitive in terms of performance, with three of them reaching >80% completeness for a mean purity of 80% down to masses of 1014 M⊙ and up to redshift z=2. Based on these results, two algorithms were selected to be implemented in the Euclid pipeline, the AMICO code, based on matched filtering, and the PZWav code, based on an adaptive wavelet approach.
The workshop on Computational Intelligence in Remote Sensing and Astrophysics (CIRSA) aims at bringing together researchers from the environmental sciences, astrophysics and computer science communities in an effort to understand the potential and pitfalls of novel computational intelligence paradigms including machine learning and large-scale data processing.
Date: July 8-14 2019
Venue: Basel, CH
Conference App will be announced on the blog.
EuroPython is an annual conference hosting ~1200 participants from academia and companies, interested in development and applications of python programming language. It's also a good opportunity for students and postdocs who wish to find a job outside academia.
For more info, contact: Valeria Pettorino
Current trends in scientific imaging are challenged by the emerging need of integrating sophisticated machine learning with Big Data analytics platforms. This work proposes an in-memory distributed learning architecture for enabling sophisticated learning and optimization techniques on scientific imaging problems, which are characterized by the combination of variant information from different origins. We apply the resulting, Spark-compliant, architecture on two emerging use cases from the scientific imaging domain, namely: (a) the space variant deconvolution of galaxy imaging surveys (astrophysics), (b) the super-resolution based on coupled dictionary training (remote sensing). We conduct evaluation studies considering relevant datasets, and the results report at least 60\% improvement in time response against the conventional computing solutions. Ultimately, the offered discussion provides useful practical insights on the impact of key Spark tuning parameters on the speedup achieved, and the memory/disk footprint.
Date: January the 24th, 2019
Organizer: Joana Frontera-Pons <email@example.com>
CEA Saclay is around 23 km South of Paris. The astrophysics division (DAp) is located at the CEA site at Orme des Merisiers, which is around 1 km South of the main CEA campus. See http://www.cosmostat.org/contact for detailed information on how to arrive.
On January the 24th, 2019, we organize the fourth day on machine learning in astrophysics at DAp, CEA Saclay.
All talks are taking place at DAp, Salle Galilée (Building 713)
14:00 - 14:30h. Machine Learning in High Energy Physics : trends and successes - David Rousseau (LAL)
14:30 - 15:00h. Learning recurring patterns in large signals with convolutional dictionary learning - Thomas Moreau (Parietal team - INRIA Saclay)
15:00 - 15:30h. Distinguishing standard and modified gravity cosmologies with machine learning - Austin Peel (CEA Saclay - CosmoStat)
15:30 - 16:00h. Coffee break
16:00 - 16:30h. The ASAP algorithm for nonsmooth nonconvex optimization. Applications in imagery - Pauline Tan (LJLL - Sorbonne Université) 16:30 - 17:00h. Deep Learning for Blended Source Identification in Galaxy Survey Data - Samuel Farrens (CEA Saclay - CosmoStat)
David Rousseau (LAL)
Machine Learning has been used somewhat in HEP in the nighties, then at the Tevatron and recently at the LHC (mostly Boosted Decision Tree). However with the birth of internet giants at the turn of the century, there has been an explosion of Machine Learning tools in the industry.. A collective effort has been started for the last few years to bring state-of-the-art Machine Learning tools to High Energy Physics.
This talk will give a tour d’horizon of Machine Learning in HEP : review of tools ; example of applications, some used currently, some in a (possibly distant) future (e.g. deep learning, image vision, GAN) ; recent and future HEP ML Kaggle competitions. I’ll conclude on the key points to set up frameworks for High Energy Physics and Machine Learning collaborations.
Thomas Moreau (Parietal team - INRIA Saclay)
Convolutional dictionary learning has become a popular tool in image processing for denoising or inpainting. This technique extends dictionary learning to learn adapted basis that are shift invariant. This talk will discuss how this technique can also be used in the context of large multivariate signals to learn and localize recurring patterns. I will discuss both computational aspects, with efficient iterative and distributed convolutional sparse coding algorithms, as well as a novel rank 1 constraint for the learned atoms. This constraint, inspired from the underlying physical model for neurological signals, is then used to highlight relevant structure in MEG signals.
Austin Peel (CEA Saclay - CosmoStat)
Modified gravity models that include massive neutrinos can mimic the standard concordance model in terms of weak-lensing observables. The inability to distinguish between these cases could limit our understanding of the origin of cosmic acceleration and of the fundamental nature of gravity. I will present a neural network we have designed to classify such cosmological scenarios based on the weak-lensing maps they generate. I will discuss the network's performance on both clean and noisy data, as well as how results compare to conventional statistical approaches.
Pauline Tan (LJLL - Sorbonne Université)
In this talk, I will address the challenging problem of optimizing nonsmooth and nonconvex objective functions. Such problems are increasingly encountered in applications, especially when tackling joint estimation problems. I will propose a novel algorithm and demonstrate its convergence properties. Eventually, three actual applications in industrial imagery problems will be presented.
Samuel Farrens (CEA Saclay - CosmoStat)
Weak gravitational lensing is a powerful probe of cosmology that will be employed by upcoming surveys such as Euclid and LSST to map the distribution of dark matter in the Universe. The technique, however, requires precise measurements of galaxy shapes over larges areas. The chance alignment of galaxies along the line of sight, i.e. blending of images, can lead to biased shape measurements that propagate
to parameter estimates. Machine learning techniques can provide an automated and robust way of dealing with these blended sources. In this talk I will discuss how machine learning can be used to classify sources identified in survey data as blended or not and show some preliminary results for CFIS simulations. I will then present some plans for future developments making use of multi-class classification and segmentation.
CosmoStat members in collaboration with colleagues at University College London (UCL) have significantly improved the analysis of dark matter maps in the Universe with new methods of data analysis (GLIMPSE method).
Managing the huge volumes and varying streams of Big Data digital information presents formidable analytical challenges to anyone wanting to make sense of it. Consider the mapping of space, where scientists collect, process and transmit giga-scale data sets to generate accurate visual representations of millions of galaxies. Or consider the vast information being generated by genomics and bioinformatics as genomes are mapped and new drugs discovered. And soon the Internet of Things will bring millions of interconnected information-sensing and transmitting devices.