summer school
of machine learning
of machine learning
20–31 august
Belokurikha Mountain, Altai Krai region
Belokurikha Mountain, Altai Krai region
|SMILES
2023
school
The Skoltech Summer School of Machine Learning (SMILES) is a 12-day intensive course dedicated to modern statistical methods of machine learning. Its objective is to bring together talented Russian youth to explore the issues surrounding artificial intelligence in the unique natural setting of the Altai Krai region, specifically in the mountainous area of Belokurikha. SMILES-2023 will cover topics fundamental to constructing contemporary predictive models in machine learning, incorporating physical and mathematical models of processes to address engineering challenges in sustainable development, ESG risk assessment, and optimization of managerial decisions for risk reduction.
Cost:
no registration fees
no registration fees
Format:
in-person and online
in-person and online
Working language:
Russian
Russian
Selection by competition:
is mandatory for both in-person and online participation
is mandatory for both in-person and online participation
venue
The summer school will take place against the backdrop of the unique views of Belokurikha Mountain. The event venue is surrounded on all sides by a mountain range. The rocks of whimsical shapes, famous springs, hikes along mountain trails with breathtaking views of the surroundings — these are just a small part of the attractions of this beautiful region.
participants
Participation in SMILES is open to master's and doctoral students from leading universities in Russia, as well as young researchers actively engaged in research in the field of machine learning and its application for modeling technical and physical systems.
selection process
Potential participants are required to submit the following set of documents:
➔ CV;
➔ Motivation letter;
➔ Brief abstract of results from one of their research projects (not more than 1 page in pdf format excluding the list of references, Times New Roman font, 12/14 pt, 1/1.5 line spacing).
It is preferable that the research topic is in the field of machine learning and data analysis or their applications, but this requirement is not mandatory.
➔ CV;
➔ Motivation letter;
➔ Brief abstract of results from one of their research projects (not more than 1 page in pdf format excluding the list of references, Times New Roman font, 12/14 pt, 1/1.5 line spacing).
It is preferable that the research topic is in the field of machine learning and data analysis or their applications, but this requirement is not mandatory.
posters
Participants who have passed the selection process must prepare and present a poster with the results described in the abstract at the conference. This is a long-standing tradition of summer schools: we want to introduce all school participants to your work. Only with a poster, you will get access to SMILES.
➔ (PDF) example of a poster made in LATEX
➔ (PDF) example of a poster made in POWERPOINT
➔ (PDF) example of a poster made in LATEX
➔ (PDF) example of a poster made in POWERPOINT
*The poster topic may not be related to machine learning and data analysis. In this case, the author must also present their thoughts on how machine learning and data analysis could allow for a more efficient solution to the problem.
moreover
Physics-Informed Machine Learning (PIML) is currently developing actively.
“
Physical sciences will receive unprecedented acceleration through the merging of modeling tools on HPC clusters and deep learning tools to enhance and accelerate this process.
directions
Related directions in machine learning considered at the summer school
probabilistic machine learning methods
machine learning methods with architectural or mathematical constraints for maintaining physical conservation laws and following trajectories of physical processes
machine learning methods that allow modeling and accounting for the geometric and topological structure of data when building predictive models
the use of predictive models for modeling physical and technical systems
training format
mandatory lectures on each direction
practical seminars mainly focused on demonstrating the operation and programming of the discussed methods and algorithms
team formation of 2–4 people and implementation of one of the proposed projects during the hackathon
session with presentations by teams on the results of the projects implemented during the summer school*
*the best presentations will be awarded with commemorative prizes
poster session with the presentation of the results of scientific works by the participants of the summer school
schedule
- Overview of machine learning methods considering physical and mathematical models of processes (Lecture)
- Introduction to the Bayesian approach to probability theory. Basics of Bayesian inference (Lecture)
- Variational inference. Doubly stochastic inference. Variational autoencoder (Lecture)
- Variational autoencoder (Seminar)
- Announcement of project topics for event participants
- Diffusion models through the prism of variational autoencoder (Lecture)
- Diffusion models. Working with formulas (Seminar)
- Stochastic differential equations. Diffusion models as a special case of SDEs (Lecture)
- Practice on diffusion models (Seminar)
- Latent models with discrete variables. Gumbel-softmax trick (Lecture)
- Discrete variational autoencoder (Seminar)
- Stochastic softmax trick (Lecture)
- Discussion of tasks where SST can be used (Seminar)
- Optimal transport. Using optimal transport and deep neural networks for generative modeling. Evaluation of barycenters based on neural networks (Lecture)
- Algorithms for calculating optimal transport (Seminar)
- Modeling manifolds (Lecture)
- Project consultations. Group work
- Neural differential equations for problems with expressed spatiotemporal dynamics (Lecture)
- Models for processing data with spatiotemporal dynamics (Lecture)
- Topological data analysis. Persistent diagrams. Estimation of deep neural network loss function properties based on persistent diagrams (Lecture)
- Efficient algorithms for calculating persistent diagrams. Use of persistent diagram features in applied tasks. Modeling 3D scans of cores, medical scans (Seminar)
- Methods of considering physical-mathematical models of processes in building predictive machine learning models. Accounting for conservation laws and equations of motion (Lecture)
- Examples of physical-mathematical models of processes in which the use of machine learning allows for accelerating calculations, assessing uncertainty. Examples of soft and hard physical constraints (Seminar)
- Assessment of the impact of climate change on physical and economic risks (Lecture)
- Modeling of atmospheric dynamics. Increasing the detail of forecasts using physically-informed models (Lecture)
- Modeling the dynamics of ocean currents and ice cover. Prediction of ice conditions (Lecture)
- Using remote sensing data to monitor the condition of the territory (Lecture)
- ML methods for processing remote sensing data (Seminar)
- Machine learning for assessing the economic impact of climate risks (Lecture, Seminar)
- Algorithms for combining heterogeneous data for modeling oil and gas fields. Approaches to implementing geological realism (Lecture)
- Deep learning models for sequential data in the oil and gas industry (Lecture)
- Presentation of reports on the results of project implementatio
- Poster session. Discussion of the results of conference participants' research
The schedule is preliminary. It is expected to include an excursion program and an increase in duration.
speakers
Evgeny Burnaev
Professor, Head of the Applied AI Center, Skoltech
Alexey Zaytsev
Ph.D., Head of the LARSS Laboratory at the Applied AI Center, Skoltech
Vladimir Vanovsky
Head of the Research Group at the Applied AI Center, Skoltech, Associate Professor at the Department of General Physics at the Moscow Institute of Physics and Technology (MIPT)
Irina Gayda
MBA, Deputy Director of the Project Center for Energy Transition, Skoltech, Independent Board Member of NOVATEK
Alexander Bernshtein
Ph.D., Professor at the Applied AI Center, Skoltech
Dmitry Vetrov
Ph.D., Professor at the National Research University Higher School of Economics (HSE)
Ilya Trofimov
Ph.D., Research Scientist, Skoltech
Alexander Korotin
Researcher at Skoltech
Nikita Lazarevich
Expert at the Applied AI Center, Skoltech
Alexander Bulkin
Senior Research Engineer at the Applied AI Center, Skoltech
Alexander Lukashevich
Research Engineer at the Applied AI Center, Skoltech
The list of speakers is preliminary and subject to change and expansion.
partners
26/JUNE
Application deadline
7/JULY
Admission lists
20/AUGUST
START >>>
registration is closed
faq
SMILES aims to unite talented Russian-speaking youth around the topic of artificial intelligence on the unique natural site of Belokurikha Mountain and online. Anyone with a personal project or publication related to machine learning can apply. This includes PhD students, master's students, industry representatives, and faculty members. We also welcome applications from candidates from other regions.
SMILES does not require any fees for participation.
The application process opens on Friday, May 12, and will be open until June 26. We will review all applications and send invitations to selected participants on July 7.
A poster is a long-standing tradition of summer school. Selected participants are asked to upload a five-minute video about their publication or project, lasting from five to eight minutes, by August 12. The video briefly explains the problem statement and main results of the work.
We want to share your work with all Summer School participants and faculty members. We have planned various activities for discussing your work, where you can also learn about the projects of other participants.
We welcome all types of projects and publications. They can include both published work and work in development.
SMILES participants will receive a digital (in the case of online participation) or paper certificate confirming their participation at the end of the event.
contacts
Any questions about the summer school can be asked in the Telegram-channel
or via e-mail at SMILES2023@skoltech.ru
Skolkovo Institute of Science and Technology (Skoltech)
Bolshoy Boulevard 30, bld.1
Skolkovo Innovation Center Territory
Bolshoy Boulevard 30, bld.1
Skolkovo Innovation Center Territory