2. ▶ From basics of Machine Learning to the cutting edge
and beyond
– Mathematical foundations;
– Real life challenges from High-Energy Physics;
– Industry-related illustrations and examples;
– Socialization and Networking.
▶ We love ML and Particle Physics ;-)
School Idea
Andrey Ustyuzhanin, NRU HSE
4. ▶ formulate a HEP-related problem in ML-friendly terms;
▶ select quality criteria for a given problem and express it in computable form;
▶ understand and apply principles of widely-used classification models (e.g. boosting,
bagging, BDT, neural networks, etc) to practical cases;
▶ design the procedure for choosing the best classifier implementation amongst a variety of
ML libraries (scikit-learn, xgboost, etc.) and deep learning architectures;
▶ understand and apply principles of generative model design;
▶ optimize features and parameters of a given model in efficient way under given
restrictions;
▶ understand optimization algorithm design and apply those to a variety of real-life
problems;
▶ define and conduct reproducible data-driven experiments.
Learning objectives
Andrey Ustyuzhanin, NRU HSE
6. ▶ Established in March 2014,
▶ Active support from industry: Yandex (more than 50 lecturers/seminarists are
Yandex employees), Kaspersky, WorldQuant, Huawei, SAS, JetBrains, and other
▶ Goal: enter the world’s top 30 Computer Science faculties by 2022
– big data storing and processing,
– software development and system programming,
– machine learning
▶ ~400 students enter bachelor program every year
▶ ~100 graduate master’s program every year
▶ Member of LHCb collaboration since June 2018
▶ Collaborates with many European and US universities
HSE Faculty of Computer Science
Andrey Ustyuzhanin, NRU HSE
7. ▶ Development and application of ML methods and tools to
Natural science challenges
▶ Close partnership with Yandex School of Data Analysis
▶ Collaborates with LHCb, SHiP, OPERA, NEWSdm experiments
▶ Research Project examples:
– Storage/Speed optimization for LHCb triggers;
– Particle Identification algorithms;
– Optimization of detector;
– Event simulation speed-up.
▶ Co-organization of ML challenges: Flavours of Physics,
TrackML
▶ Open for interns, graduate students and post doc candidates!
Laboratory of Methods for Big Data Analysis
Andrey Ustyuzhanin, NRU HSE
hse_lambda
8. ▶ 2015 – Academic University, Saint Petersburg, Russia
▶ 2016 – Lund University, Sweden
▶ 2017 – Imperial College London, Reading, UK
▶ 2018 – Oxford, UK
▶ 2019 – Hamburg, Germany
▶ 2020 – EPFL + Online
▶ 2021 – EPFL + Online
MLHEP - iterative education experiment
Andrey Ustyuzhanin, NRU HSE
12. ▶ Main activities:
– lectures, seminars, competitions
▶ Most of the lectures are pre-recorded, Q&A are part of the
seminar
▶ Schedule
– https://indico.cern.ch/event/1025052/timetable/ , CEST Time Zone
▶ All classes are going to be recorded and stored for the
participants at ALEMIRA platform
Recap
Andrey Ustyuzhanin
13. ▶ ECTS credits (MLHEP as EPFL official course)
– 4 Credits
▶ Evaluation metrics:
– Quizzes – 1 score point each quiz
– Seminar tasks – 6 score points if all tasks performed correctly
– Data challenge & presentations – wait for Nikita’s presentations
– Watching videos and interactive contents also adds to the metrics ~0.1 of points
▶ Evaluation criteria
– 60% + of top score
– Will be calculated upon the end of the school (~2nd
of August)
▶ Certification
– Certificate of excellence will be awarded with the same criteria as for ECTS credits
Qualification
Andrey Ustyuzhanin
14. ▶ Social Activities, past years examples
– Zoom lunch with lecturers
– TED-style talks
– Mafia-style games
– 3D-shooter teamplay online
▶ This year ideas:
– Random coffee, around 26th
of July
– “Among Us” tournament
▶ Communication channels:
– Zoom for interactive classes (seminars, live talks)
– MS teams for Q&A, technical assistance, social activities
Social activities
Andrey Ustyuzhanin