Data mining, machine learning and deep learning

Part I (main lecture) : Machine learning and deep learning.

Adeline Paiement, Université de Toulon

Data mining and machine learning have universal applications for extracting meaningful information from large and complex datasets. The aim of this course is to introduce and familiarise students of various backgrounds to these tools and to their possible applications in their own disciplines. The first part of the course will present the main concepts of data mining and machine learning, from data acquisition and preparation to the various possible analyses. The second part of the program will focus on the deep learning paradigm that attains state-of-the-art results in many fields. The theoretical grounds for the methods will be introduced and complemented by examples from a variety of disciplines, mainly within the physics and medical fields.

Part II (~2hrs) : Examples in astrophysics with a focus on the measurement of photometric redshifts.

Kai Polsterer, HITS gGmbH, Heidelberg, D]

Astronomical surveys are gathering huge amounts of data (the near-future LSST will observe half the sky every three days at several wavelengths). The recession velocities, known as redshifts, of the millions of galaxies in these surveys serve to characterize the evolution of structure in the universe, with implications on fundamental issues such as dark matter.  Machine learning now makes it possible to estimate redshifts directly based on images, while automatically accounting for the fact that galaxies have different shapes and colors, and the images are contaminated by various sources of noise and defects.

Part III (~2hrs) : Examples in particle physics at CERN.

Maurizio Pierini, CERN, CH

Particle physics is another field of science that deals with avanlanches of data. Deep learning technologies are being developed for various applications. An overview of some of these will be presented, with a focus on one to gain insight into some of the tricks of the trade.

Public : PhD students of ED-182

Other PhD students in sciences, motivated Master’s students, researchers

Dates : 2019 – May 21, 22, 23, 24

May 21 : 9h-12h30 + 14h-16h00 (with a break) – AP and KP

May 22 : 10h-12h + 14h-16h00 (with a break) – AP and KP

May 23 : 10h-12h – AP

May 24 : 14h-16h (with a break) – MP

Place (TBC) : Strasbourg astronomical Observatory [+ IPHC Cronenbourg on 24/5 if convenient]

Introduction to Numerical Simulations and High Performance Computing: From Materials Science to Biochemistry

Mauro BOERO, Guido ORI

IPCMS

Advances in the theoretical methods and high-performance computing (HPC) platforms make nowadays numerical approaches a versatile, reliable and precise tool for performing virtual experiments, complementing the real ones, and providing access to details escaping experimental probes. This has paved the route to an entire new multidisciplinary field where design of materials with specifically tuned properties, advanced drug discovery or proteins and nucleic acids interactions represent just a few examples of the possibilities disclosed. Given this scenario, knowledge of the basic principles, their strength and also their limitations has become a fundamental background for researchers working in the fields of materials science, chemistry and biology. The present course is aimed at providing a general introduction to the most advanced algorithms and methods in numerical simulations to students with a standard background in classical and/or quantum mechanics and thermodynamics. Theoretical approaches are presented and their analytical formulation discussed in view of their numerical (discrete) version required for any implementation in a computer code. Examples of applications are given, providing an insight into the type of information that can be extracted from these numerical approaches. Finally, an overview of the HPC architectures available to date and in the near future, along with their exploitation by parallel programming tools will be presented as a necessary complement to the basic theory developed for an efficient use of the continuously evolving technologies in massively parallel computing.

Main topics:

1. Classical Molecular Dynamics (MD):

Newtonian and Lagrangean dynamics: Basic algorithms

Time averages and ensemble averages

2. First Principles MD:

Brief review of Hartree-Fock approaches and Density Functional Theory

First Principles Molecular Dynamics (FPMD)

Mermin functional and Free Energy MD (FEMD)

3. Hybrid schemes:

Combined classical MD and DFT-based MD: Hybrid QM/MM methods

4. Advanced Methods: Reactive schemes

Free Energy sampling techniques and reaction path sampling

5. Advanced Methods: Path Integral and non-Adiabatic Methods

Quantum nuclei and surface hopping dynamics

6. Brief overview of parallel architectures and parallel programming

Time:    January 22nd, 24th, 29th and 31st; and February 5th and 7th 2019 from 2pm to 4 pm (12 hours) 

Place:   Auditorium, Institut de Physique et Chimie des Matériaux de Strasbourg, 23, rue du Loess, 67034 Strasbourg

QMat Thesis Award

1) Study of secondary neutron production in accelerators (D. Husson)

2) Integration of the veto to muon and analysis of its performance on the cosmogenic background noise of the neutrino JUNO experiment (M. Dracos)

3) Supersymmetric particle search in the CMS experiment with LHC Run 2
data (C. Collard & E. Chabert)

Calendar of the doctoral contract award competition EUR 2018 :

until July 30 at 12 noon, deadline: receipt of applications (only complete applications will be considered), until August 8, 2018 : Candidates authorised to continue the competition must submit a document describing the Master’s internship in the form of a pdf file (2 pages) to the Director of the Doctoral School Prof. Aziz Dinia (aziz.dinia@ipcms.unistra.fr).

The selected candidates will be informed later and will be invited to the auditions which will take place from August 25, 2018. Candidates must carry a national identity card or passport. Candidates will give a presentation on their master’s internship (10 minutes) and answer questions from the jury. For the ranking in the competition, equal importance is given to the audition and the file.

Submit :

Form in English

Bourse de thèse QMat

1) Etude de la production des neutrons secondaires auprès des accélérateurs (D. Husson)

2) Intégration du véto à muons et analyse de ses performances sur le bruit de fond cosmogénique de l’expérience neutrino JUNO (M. Dracos)

3) Recherche de particules supersymétriques dans l’expérience CMS avec les donnéesdu Run 2 du LHC (C. Collard & E. Chabert)

Calendrier du concours d’attribution du contrat doctoral EUR 2018 :

• jusqu’au 30 juillet à 12h, date limite : réception des candidatures (seules les demandes complètes seront considérées),
• jusqu’au 8 août 2018 : Les candidats autorisés à poursuivre le concours doivent soumettre un document décrivant le stage de master sous la forme d’un fichier pdf (2 pages) au directeur de l’École Doctorale Prof. Aziz Dinia (aziz.dinia@ipcms.unistra.fr)
• les candidats retenus seront informés ultérieurement et seront invités aux auditions qui se dérouleront à partir du 25 Août 2018. Les candidats doivent se munir d’une carte nationale d’identité ou d’un passeport. Les candidats font un exposés sur leur stage de master (10 minutes) et répondent aux questions du jury. Pour le classement au concours, une importance égale est attribué à l’audition et au dossier.

Déposer une candidature :

Formulaire de candidature

Starting 24 Mai 2018 at Ipcms, Prof. J-P Bucher

Clusters and Nanoparticles: the Essentials