[Todos] PhD opening at IJCLab-Orsay : Fair Universe : uncertainty-aware Artificial Intelligence for ATLAS experiment Higgs boson Physics

Lun Nov 28 12:34:08 -03 2022


En 28 nov. 2022 15:59, en 15:59, David Rousseau <rousseau en ijclab.in2p3.fr> escribió:
>**
>
>*Hi, *
>
>*
>
>Please find the PhD opening below. Thanks for forwarding it to 
>potentially interested M2 students.
>
>Best regards,
>
>David Rousseau
>
>title : Fair Universe : uncertainty-aware Artificial Intelligence for 
>ATLAS experiment Higgs boson Physics
>
>
>Application deadline 11 December 2022.
>
>
>This 3 years Université Paris-Saclay PhD is funded through Fair
>Universe 
>project, 
>https://cs.lbl.gov/news-media/news/2022/new-fair-universe-project-aims-to-build-supercomputer-scale-ai-benchmarks-for-he
>
><https://cs.lbl.gov/news-media/news/2022/new-fair-universe-project-aims-to-build-supercomputer-scale-ai-benchmarks-for-hep-and-beyond/>
>
>p-and-beyond/ 
><https://cs.lbl.gov/news-media/news/2022/new-fair-universe-project-aims-to-build-supercomputer-scale-ai-benchmarks-for-hep-and-beyond/>
>
>The PhD will be co-supervised by David Rousseau (IJCLab, CNRS/IN2P3, 
>ChaLearn) and Isabelle Guyon (Google Brain, UPSaclay/LISN, ChaLearn).
>
>It is open to students expecting to obtain a Master 2 (or equivalent)
>in 
>Particle Physics by summer 2023. It would typically start in Oct 2023 
>and possibly be preceded by an internship in late spring 2023. The 
>successful candidate is expected to have an excellent track record in 
>physics, but also to demonstrate interest in advanced Machine Learning 
>algorithms.
>
>
>Physicists ultimately write papers with measurements which always 
>include an assessment of uncertainties, including the impact of
>Nuisance 
>Parameters (the parameterized unknowns) in systematic uncertainties.
>How 
>to evaluate the uncertainties on a more and more complex model? How to 
>build confidence in a complex ML model, and how to convince our peers 
>about it? How to deal with the uncertainties on the input of the models
>
>to maximise the overall accuracy?
>
>
>The Fair Universe project aims at releasing public datasets of 
>increasing complexity and at developing strategies to evaluate 
>uncertainties and reduce them with Machine Learning, in the context of 
>particle physics.
>
>
>The PhD candidate will be a member of the ATLAS experiment.
>
>In the first year, s/he will work at half-time on an ATLAS
>Qualification 
>Task to be defined, and on the other half time on benchmarking
>different 
>Machine Learning uncertainty mitigation techniques on public datasets 
>developed by collaborators in Fair Universe project, starting from the 
>techniques available in the literature (see bibliography). S/he will 
>participate to the definition of public competitions on the theme.
>
> From the second year on, s/he will develop an analysis of the 
>measurement of Effective Field Theory parameters for the Higgs boson in
>
>the off-resonance Higgs to four lepton channel, using advanced Machine 
>Learning algorithm. In particular, Simulation Based Inference  (SBI) 
>allows for fully exploiting the event by event information by
>estimating 
>the likelihood ratio with a Neural Network; it has been demonstrated to
>
>be particularly efficient in this channel due to a quantum interference
>
>effect between the Higgs signal and the main background. The improved 
>sensitivity allows unique constraints on several Effective Field Theory
>
>operators from LHC Run 3 dataset.
>
>Currently, the impact of Nuisance Parameters on the sensitivity is 
>evaluated a posteriori. The techniques developed during the first year 
>will be applied (and possibly complexified) on this channel, 
>anticipating that different types of Nuisance Parameters might call for
>
>different strategies.
>
>
>David Rousseau is a particle physicist member of the ATLAS 
>collaboration. After many years developing the ATLAS experiment 
>software, he has been promoting the use of Machine Learning in High 
>Energy Physics since 2014  with developments in particular concerning 
>generator models or Simulation Based Inference. Within ChaLearn and
>with 
>Isabelle Guyon he has organized the Higgs Boson Machine Learning 
>challenge in 2014, which has played a major role in promoting Machine 
>Learning to physicist and raise awarenesss about Machine Learning in
>the 
>High Energy physics community. They have also organized the TrackML 
>challenge to help the development of advanced algorithms for particle 
>tracking at the LHC.
>
>
>Isabelle Guyon is director of research at Google Brain, in detachment 
>from her position as professor of Artificial Intelligence at Université
>
>Paris-Saclay (Orsay). She specializes in data-centric AI, statistical 
>data analysis, pattern recognition, and machine learning. Her areas of 
>expertise include computer vision, bioinformatics, and power systems. 
>Her recent interests include meta-learning and applications of machine 
>learning to the discovery of causal relationships. Prior to joining 
>Paris-Saclay she worked as an independent consultant and was a 
>researcher at AT&T Bell Laboratories, where she pioneered applications 
>of neural networks to pen computer interfaces (with collaborators 
>including Yann LeCun and Yoshua Bengio) and co-invented with Bernhard 
>Boser and Vladimir Vapnik Support Vector Machines. She is president of 
>Chalearn, a non-profit dedicated to organizing challenges.
>
>
>The PhD candidate will be based at IJCLab-Orsay, with frequent short 
>stays at CERN and participation in relevant workshops and conferences. 
>IJCLab is the largest particle physics laboratory in France, depending 
>mainly on Université Paris-Saclay and CNRS/IN2P3. It is one of the 
>founding laboratories of ATLAS (then with the name LAL) with major 
>contributions to the Liquid Argon Calorimeter and to software, and more
>
>recently to the Inner Tracker and High Granularity Timing Detector, and
>
>with major contributions to Higgs boson physics.
>
>
>Interested candidates should provide a cover letter,  CV and 2 letters 
>of references.  They should not hesitate to contact David Rousseau 
>and/or Isabelle Guyon for details before submission, in particular if 
>they are not already in the French system.
>
>*
>
>-- 
>-----------------------------------------------------------------------------------------------------
>      David Rousseau,  DR CNRS,david.rousseau en ijclab.in2p3.fr      
>              https://users.ijclab.in2p3.fr/david-rousseau/       
>    IJCLab-Orsay,  CNRS/IN2P3, Université Paris-Saclay, France
>   @ IJCLab   :  Bat 200 Pièce 142 : +33 (0)1 64 46 85 91
>   @ CERN :  40/1-D12                : +41 (22) 76 73857
>-----------------------------------------------------------------------------------------------------
------------ próxima parte ------------
Se ha borrado un adjunto en formato HTML...
URL: <http://mail.fisica.unlp.edu.ar/pipermail/todos/attachments/20221128/85fe221e/attachment-0001.html>