• A
  • A
  • A
  • ABC
  • ABC
  • ABC
  • А
  • А
  • А
  • А
  • А
Regular version of the site

'We Pursue Whatever Interests Us!'

MIEM HSE Professor Lev Shchur is celebrating his 70th birthday. In his recent interview with MIEM HSE News Service, he talks about the first scientific computer network created in the Soviet Union, the state of computational physics today, and the current online and offline work being conducted by his research team of undergraduates and faculty members.

Lev Shchur, MIEM HSE Professor, Doctor of Physics and Mathematics and prominent Russian computational physicist, is celebrating his 70th birthday. Dr Shchur is the leading research fellow at the RAS Landau Institute of Theoretical Physics, Head of the Department of Applied Network Research at the Chernogolovka Scientific Centre, and creator of Russia's first scientific computer network using fibre-optic communication.

At HSE University, Dr Shchur supervises the Master's Programme in Supercomputer Simulation in Science and Engineering—in addition to master's students, the programme includes bachelor's and doctoral students majoring in computational physics.

On behalf of MIEM HSE faculty and students, we wish Dr Shchur a happy birthday and many new achievements and insights! On the eve of his anniversary, Dr Shchur kindly agreed to answer a few questions from the MIEM News Service.

—You created the first scientific computer network built in Russia using fibre optic communication. In fact, it was the first scientific corporate internet before the word 'internet' was even known. How was this idea conceived and brought to life?

—It all began in 1989, when our team at the Landau Institute for Theoretical Physics started corresponding by email with our foreign colleagues. By that time, we already had an internal network, or rather two networks: DECnet running on a VAX 11/780 cluster and another network connecting IBM PCs through a makeshift gateway.

Email messages sent from different terminals were collected on one server and forwarded to the X25 network provided by the Kurchatov Institute for communication with the outside world; email was sent out several times a day

Although the process took considerable time, email correspondence facilitated exchanges among scientists and made it possible to promptly arrange regular and reciprocal academic visits.

After I spent quite some time doing research in Italy in 1991, I came to realize the benefits of connecting remote computers online and decided to create this possibility at my institute in Russia.

It was clear that future development would involve TCP/IP, where IP stands for 'Internet Protocol', which eventually gave the global network its current name    

Developing this new technology was an exciting process that would add a new dimension to my work for years to come. Building a fibre-optic network to connect a number of scientific institutions in Chernogolovka was our next step. I am grateful to Academician Osipyan, who directed the Scientific Centre in Chernogolovka at the time, for his confidence in our ability to do so, and to Academician Fortov, Chairman of the Russian Foundation for Basic Research (RFBR), for approving the project’s funding. During a crucial discussion on November 7, 1993, I showed my senior colleagues how the internet could be used for interactive work. At that time, all browsers were alphanumeric (consisting of letters and numbers). It was only towards the end of that year that the first graphical browser appeared in a beta version. There were just 50 websites worldwide – we were making our first steps in the virtual space. Indeed, email was considered the ultimate form of information technology at the time, and few people saw the potential for anything beyond it. This was how Chernogolovka became an internet leader in Russia.

Professor Lev Shchur, MIEM
Professor Lev Shchur, MIEM

—You specialize in computational physics, which is about using powerful computers to solve complex problems. What are some of the problems tackled by computational physicists today?

—A computational physicist works at the intersection of applied mathematics and specific fields of physics. They need to have in-depth knowledge of physics, mathematical methods, and computer science to work on a wide range of problems. Computational physics is a third scientific approach that complements experimental and theoretical methods. Plus, a fourth approach, which is based on extracting new knowledge from big data, has emerged in the last five years.

—You have mentioned the third and fourth approaches, and the first two are...

—Theoretical and experimental!

MIEM master's programme and your work in particular bring together a large number of students and research teams. What areas are being addressed by student research and how is it being managed? Could you give a few examples that you find particularly interesting?

—The context of student research can be described as an informal computational physics laboratory, and our current work covers several subject areas. In statistical mechanics, we work to develop effective algorithms for computer simulation of phase transitions in classical and quantum mechanics. While real-life experiments require new instruments and measurement approaches, computer-based experiments call for high-performance supercomputers and new computational methods. We laid the groundwork for a supercomputer centre at HSE to create the technical capacity and have been working to generate new ideas and to design new approaches, such as efficient parallelisation methods to maximize supercomputer performance. The master’s programme in supercomputer simulations aims to build the required skills.

In recent years, the master’s programme has been expanded to include topics such as deep machine learning and big data processing

The master’s programme uses a project-based approach and individual development plans for learners. For example, my colleagues and I are currently supervising the research work of two dozen undergraduate and doctoral students.

They are working on different research topics, for example, the flow of complex fluids in microchannels of complex geometry with reference to microfluidics and transport of particles in blood vessels, magnetic properties of foam structures that show good potential as hydrogen reservoirs for a hydrogen economy, designing controlled methods to determine the density of states for statistical models, direct measurement of entropy in studying multi-critical points on phase diagrams of substances, and using deep machine learning to explore the characteristics of phase transitions.

Last year, we launched a project to study the properties of hard alloys using machine vision and deep learning. Another successful project undertaken with bachelor's students uses statistical physics methods to study social and economic models. Work is underway in areas such as physical and chemical systems, quantum computing, and elements of quantum computers.

In addition, we joined my colleague and former student from the Kurchatov Institute to launch a project on using NLP (Natural Language Processing) with elements of AI to process scientific information. There are no limits to the range of topics we can address in our research apart from the need for it to be relevant and advanced, supported by scientific foundations, and for findings to be published in leading scientific journals.

We pursue whatever interests us!

My colleagues and I have developed a multi-level system of working with students.

We observe students starting from their first year and offer them opportunities to engage in research. Once a week, we hold meetings attended by some three dozen faculty members, as well as doctoral and master's students

The main purpose of these meetings is for everyone to update the group on the status of their research. These brief updates are followed by more detailed discussions in Telegram chats among faculty and students working on specific projects and assignments. Such in-depth discussions are held as needed. Each Wednesday, 90-minute Computing Environments seminars are held to discuss new ideas and findings. Both our students and researchers from other Russian and foreign scientific institutions attend these sessions. And finally, we convene HPC (high-performance computing) workshops about once a month, which feature updates on current issues and achievements by researchers from HSE and other Russian academic institutions.

All of the above serves to create an atmosphere of collective scientific work.

—Working with big data requires mastery of new tools. Lately, the term 'supercomputer' has been included in the title of the master's programme. What kinds of specialist knowledge and skills are required to operate high-performance computers?

—First, one should have basic mathematic knowledge, which is acquired during the first three years at MIEM. Other things to learn include the logic of programming languages, hybrid computing architecture, and the ability to find and fix errors in code. A desire to learn and create new things is also important.

—Given the changes taking place in Russia's higher education system, what would you describe as the key challenges to higher education in the STEM fields?

—In my opinion, most changes serve the modern trend of packaging a product nicely to make it competitive. Back in the 1980s, Academician Larkin, a prominent theoretical physicist in the field of quantum solid-state physics, made a remark at a Scientific Council meeting on modernisation of higher education. He said, 'Does this mean that the old-fashioned way of making children no longer works?' As the founder of one of the world's best schools in theoretical physics, he had every right to say so.

I believe in evolutionary development of education where change is an organic process originating from within the academic team and school

Every school has its own unique character and features, and by getting to know a number of different universities, one can see how diverse they really are. Different approaches to learning and research management at universities are often a topic of conversation during after-session hours at international conferences.

In general terms, a bachelor's course should equip students with a solid foundation of fundamental knowledge. Building on this knowledge, a master's course should further teach students new methods, approaches, algorithms, and platforms for operating cutting-edge IT and computing tools. In each specific subject area, its current developments should be discussed and fed into the learning process.

In our research course for master's students, we make sure to study recent papers published in the last year or so. The only exception is made for fundamental studies which are important for understanding more recent trends.

—We have lived through two challenging years of the pandemic. With universities gradually moving back to an in-person mode, we now see that it can be as much of a struggle, if not more so, compared to the earlier switch to online classes. Many students and faculty alike are now used to working online. What is your perspective, and which mode do you prefer?

—We first experimented with video conferencing back in 2008 using a Video Grid system built at the RAS Innovation Centre in Chernogolovka and using guidance from Argonne National Lab experts led by Professor Ian Foster, co-creator of the Grid architecture. This was a multi-server, multi-screen, and multi-window system, and we used it to conduct international seminars, including a seminar held every month with colleagues from the leading European Jülich Supercomputing Centre in Germany. For my colleagues and me, transitioning to an online mode was a natural extension of the already familiar method of organizing and managing scientific research.

But I don’t embrace the online mode in the educational process. It’s important for me to see a sparkle in my students' eyes, to follow their reaction to what they have heard, and to take their questions. Finally, learning is something that takes place in a team environment

Seminars, group meetings and discussions are integrated into our research work. Research is a collective process. Even mathematicians need to network, although much of their work is done in solitude. The learning process is also collective. The online mode hinders student-to-student interaction, which experienced teachers can easily facilitate during an in-person session. Strong students come from strong groups. There are virtually no cases of strong students coming from weak groups.

On the other hand, the online mode allows us to expand the range of potential speakers at scientific seminars, and over the last two years, we have had guest speakers from other cities and countries. The online mode is also preferable for group meetings, because it saves time and accommodates a larger audience.

—Your responsibilities include teaching, overseeing the master's programme, serving as an academic supervisor and working for the Russian Academy of Sciences and a number of major scientific institutions. How do you make time for everything? And what advice would you give to your students?

—The most important thing is to really love your work – your studies are your work! – and to be respectful towards your colleagues and students.

Author: Oleg Myslyuk

April 21, 2022