This year, besides being as always available to be a supervisor for Ph.D. students interested in research in particle physics with the CMS experiment at the CERN LHC, as well as for general studies of applications of advanced Machine Learning techniques to particle physics problems (see at the bottom for three available lines of research which I offer to supervise students on), I have funded a special position among the 23 above for a joint Ph. D. with the Université Clermont Auvergne, focusing on cutting-edge research at the crossroads of particle physics and computer science.
The recipient of this grant is expected to spend half of his or her time in Padova and the other half in Clermont-Ferrand (except for conferences, trips to CERN and other research nodes, etc.), and be co-supervised by Prof. Julien Donini and myself. Because of the period in France, the grant amount is higher than the other positions offered in the call. But the reason why this position is attractive is because it is focused on a research plan that Prof. Donini and myself have recently undertaken, to exploit cutting-edge deep-learning techniques for the design of new particle physics experiments. This is a largely unexplored area of research, made viable only recently by advancements in differentiable programming techniques.
The MODE Collaboration, which I founded and lead, includes an outstanding pool of physicists and computer scientists who collaborate with the aim of exploiting differentiable and probabilistic programming in the optimization of measurement instruments that base their functioning on the interaction of radiation with matter.
The above loose definition allows to include detectors for high-energy physics experiments, such as those we will design for future colliders, as well as smaller, more specialized instruments for astro-particle physics applications, as well as more industrial-leaning devices such as muon tomographers and proton imagers for proton therapy facilities. You can read a short summary of the MODE research plan in this recent article.
MODE includes the following people and institutes, with which the Ph.D. candidate will have a chance to interact and cooperate:
– at the INFN, Sezione di Padova: Pablo de Castro Manzano, Tommaso Dorigo, Mia Tosi, Giles C. Strong, Lukas Layer, Hevjin Yarar;
– at CERN: Jan Kieseler
– at Oxford University: Atilim Gunes Baydin
– at Université de Liege: Gilles Louppe
– at New York University: Kyle S. Cranmer
– at Université catholique de Louvain: Christophe Delaere, Andrea Giammanco, Pietro Vischia
– at Université Clermont Auvergne: Julien Donini
– at National Research University Moscow: Denis Derkach, Fedor Ratnikov, Andrey Ustyuzhanin
The position includes a small allowance for travel to conferences and research-related activities, plus extra travel funds recycled from remainders of the AMVA4NewPhyscs ITN. I expect that the recipient will learn a lot of physics and machine learning by interacting with a stimulating collaborative environment. Two of the MODE projects have in fact just started, and we are laying the grounds for groundbreaking research activities. I am of course available to tell you more about these if you contact me (my email is myfirstname.mylastname (at) gmail.com).
What are the competences you should possess to be a good match for this position? You should have a masters degree in Physics or Computer Science (Statistical Sciences could also work), and some familiarity with machine learning and data analysis. Knowledge of particle physics is useful but not a mandatory requirement.
Finally, since I am offering myself as a mentor of brilliant physics and computer science students here, I feel it is a good idea to point you to my personal web page, where you may find information on my research and mentoring activities, among a lot of other material related to my scientific output and initiatives.
——————————-
Concerning the other lines of research which you may pick as your PhD topic if you are selected for a Ph. D. in Physics at the University of Padova and want to work under my supervision and within my research group: besides the one I described above, here are the topics I suggest. Because these themes are not connected with a special selection like the joint PhD one discussed above, if you win a position we can certainly remodulate these topics to fit better your specific interests.
1) Inference-aware dimensionality reduction for precise Higgs boson measurements: Collider physics measurements rely on a massive reduction of the measurement of millions of electronic signals into few, or even a single, variable capable of characterizing the originating collision as belonging to a process of interest, or to measure some physical quantity. This dimensionality reduction loses much of the relevant information, because it is performed without directly optimizing the quantity of interest – the smallest uncertainty on the final measurement. Using the INFERNO deep learning technique this information loss can be minimized. We propose to use it to improve the measurement of systematics-ridden Higgs boson properties with Run 2 CMS data. The PhD student will be embedded in the CMS-Padova research group, become an author of CMS publications, and participate to the CMS research with frequent visits at CERN.
2) Unsupervised Learning for New Physics Searches at the LHC: Searches for new physics phenomena in high-energy particle collisions are typically performed by the LHC experiments by considering a new physics model in guiding the data selection and information extraction procedures. In so doing, theoretical models can be excluded effectively; yet the large parameter space of potential new physics signatures remains largely unexplored. A model-independent approach is proposed to fill that gap, to perform a search in CMS data based on novel unsupervised learning algorithms that may detect anomalous regions of the parameter space. The PhD student will be embedded in the CMS-Padova research group, become an author of CMS publications, and participate to the CMS research with frequent visits at CERN.
3) Deep Learning Approach to Hybrid Calorimetry: Differentiable programming and probabilistic programming techniques allow to explore new solutions to the problem of precisely measuring the interaction of radiation with matter, by e.g. combining tracking and calorimetry into a hybrid detector and finding innovative, entirely new design solutions to the maximization of experimental resolution and to the geometry of apparata. In this work we will explore those solutions for applications to future colliders as well as astrophysics experiments. This work will be performed within the MODE collaboration, in cooperation with researchers in physics and computer science from CERN, New York University, Oxford University, Univ. Clermond Auvergne, Univ. Cath. de Louvain, Univ. Liege, and HSE Moscow.
—
Tommaso Dorigo (see his personal web page here) is an experimental particle physicist who works for the INFN and the University of Padova, and collaborates with the CMS experiment at the CERN LHC. He coordinates the MODE Collaboration, a group of physicists and computer scientists from eight institutions in Europe and the US who aim to enable end-to-end optimization of detector design with differentiable programming. Dorigo is an editor of the journals Reviews in Physics and Physics Open. In 2016 Dorigo published the book “Anomaly! Collider Physics and the Quest for New Phenomena at Fermilab“, an insider view of the sociology of big particle physics experiments. You can get a copy of the book on Amazon, or contact him to get a free pdf copy if you have limited financial means.
More Stories
Trump Goes Full Bonkers, Attacks the Constitution
A Guide to Caring for a Large Dog
My aims for next year (2022-23) – Mr. Hill’s Musings