TRIUMF Canada's National Laboratory for Particle and Nuclear Physics STUDENT JOB PROGRAM Summer 2020 job posting Job number TR20-2-7 | ||
---|---|---|
Position Title:Junior Data Scientist | ||
Name of Project:Deep Learning for Event Reconstruction in Water Cherenkov Detectors | ||
Overview:Neutrino oscillations are the only experimentally verified observation of a process not described by the Standard Model of particle physics. Since their discovery in 1998 we have learnt a lot about the parameters that govern these oscillations, but there are still many questions remaining. Perhaps the most exciting of these is to determine whether neutrino oscillations violate charge-parity (CP) symmetry, and so could potentially explain why we live in a matter-dominated universe.The T2K experiment and Nobel prize-winning Super-Kamiokande experiment are powerful tools for exploring neutrino oscillations. The next generation Hyper-Kamiokande experiment, coming online in the middle of the next decade, has significant discovery potential for CP violation, as well as other phenomenon such as proton decay, supernovae and other multi-messenger astronomical events, and dark matter. The Canadian group is working on a number of interesting R&D efforts aimed towards maximizing the potential of Hyper-K. A background and a source of major systematic uncertainty in the CP violation analysis is caused by events where a gamma ray is produced in a neutrino interaction instead of the usual charged lepton (electron). The goal of this project is to develop a full data analysis chain for the measurement of this so called Neutral Current gamma (NCgamma) background in the intermediate (distance) water Cherenkov detector (IWCD) of the Hyper-K experiment. Supervised learning techniques such as Convolutional Neural Networks and perhaps more advanced models such as Graph Neural Networks and PointNets will be applied and tuned. The output of these networks will be implemented in a NCgamma interaction rate (cross-section) analysis including evaluation of statistical and simplified systematic uncertainties. Time permitting, generative deep learning methods can be applied to limit the impact of systematic effects. Such methods will include Generative Adversarial Networks (GANs) and their variants such as CycleGANs. Approaches with alternative generative methods such as Variational Auto Encoders (VAEs), will also be considered. The successful completion of the project will have a substantial impact on key scientific goals of both the Super-K and Hyper-K experiment, such as the CP violation measurement and other areas of particle physics and astronomy. | ||
Duties:Major duties include:- Development and tuning of deep learning models for analysis of water Cherenkov data. - Development and deployment of statistical data analysis tools. | ||
Skills learned during this work experience:- Experience in many aspects of a complicated physics problem. This will range from low level understanding of photosensors to the description of the high level capabilities of a next-generation neutrino experiment.- Machine learning concepts and experience with deep learning libraries such as pytorch - Experience with generative deep learning methods - Critical thinking and problem solving - Presentation skills from reports in regular meetings at national and international levels | ||
Qualifications:We are looking for a motivated physics student to work on an exciting high-energy physics project. The core skills that we are looking for include:- Strong competence in the use of computational methods for data analysis. - Ability to quickly learn new software tools and software packages. - Experience with using Linux OS. - Python programming experience. - Experience with numerical libraries numpy, matplotlib - Understanding of machine learning and deep learning concepts. - Experience with deep learning and machine learning frameworks e.g. pytorch, scikit-learn - Senior undergraduate level knowledge of Quantum Mechanics and electromagnetism - Ability to independently solve problems and investigate alternatives. Beneficial skills would include: - Understanding of the basics of a particle physics detector. - Understanding of concepts in probability theory and statistics. - C++ programming experience - Experience with batch processing systems (slurm, torque) - Experience with versioning tools (git) | ||
Shiftwork required:No | ||
Period of work:May-August 2020 with possible 4-month extension | ||
Salary is commensurate with academic progress and previous relevant work experience, and
ranges from $ 2080 to $ 2800 per month plus 4% vacation pay. TRIUMF pays round-trip airfare (this does not apply to Vancouver/Victoria students); for Vancouver Island students, TRIUMF will pay ferry costs.
TRIUMF is an equal opportunity employer committed to diversity in the workplace, and we
welcome applications from all qualified undergraduate students as defined below:
Applications must be received at TRIUMF by 4:00pm Pacific time on 2020-01-26. To apply for any of the job postings, you must submit one application for each and every job for which you wish to be considered. Please combine all documents into one PDF for each job. All applications can be attached to one email in PDF format. Please do not send electronic documents in formats other than PDF. Please save the PDF as lastname-job number (eg Smith-15.pdf)
Students within an university co-op program MUST apply for TRIUMF jobs through their university co-op education office.
No phone calls please. TRIUMF wishes to thank all applicants for their interest, and regrets that only those being considered will be contacted.
For more information, visit our web site http://www.triumf.ca |