How Programmers Help Unravel the Secrets of Physics

The Summer School on Machine Learning in High Energy Physics, which was co-organised by the HSE Faculty of Computer Science and Russian internet company Yandex, has ended. Below, the Head of HSE's Laboratory of Methods for Big Data Analysis, Andrey Ustyuzhanin, talks about the various ways in which physicists and programmers cooperate. He also discusses how researchers from HSE and Yandex have been participating in CERN experiments and how ordinary smartphone users can help unravel the secrets of the universe.

How Machine Learning Helps Physicists

Anyone even marginally interested in science or familiar with the television series The Big Bang Theoryhas probably heard of the European Organization for Nuclear Research (CERN) and the Large Hadron Collider (LHC). But not a lot of people know that scientists from Yandex, and now HSE – more specifically, the research group headed by Andrey Ustyuzhanin – has been working with CERN since 2011. This is all because physicists are unable to avoid searching for elementary particles in their research without the help of Big Data experts. ‘Machine learning is used in practically all fields being developed by CERN specialists,’ Ustyuzhanin comments.

What is more, machine learning's various applications are growing constantly. ‘We began using our technologies to optimize network-attached storage,’ Ustyuzhanin continues, referring to CERN’s datacentre, GRID. ‘Roughly speaking, data can be stored using fast but expensive disks or cheap but slow tapes. Using statistics on file access, we predict what can be left on disks and what should be saved on tapes. As a result of such a model, you can save up to 40% on expensive disks without compromising the speed at which the entire system operates,’ he says.

This model is currently being used on part of CERN's grid, which belongs to the LHCb experiment. The main goal of the experiment is to study matter-antimatter asymmetry in the decay of particles known as b-quarks. In July 2015, LHCb announced that it had discovered a new class of particles called pentaquarks. The Yandex School of Data Analysis (SDA) became an LHCb participant in 2014.

What They Say About the Summer School

According to Andrey Ustyuzhanin, the Summer School on Machine Learning in High Energy Physics (MHLEP-2015) was ‘a natural continuation of the laboratory's relationship with CERN.’ ‘As for physics, we invited four scientists from various CERN experiments who talked about how they use machine learning to analyse data, check physics hypotheses, search for new theories, and partake in online event selection. At the data handling stage, 99.9% of events are “pitched aside,” and it’s really important that there not be significant physics-related data in this apparent debris,’ Ustyuzhanin notes.

There were twice as many applications for the Summer School as the event's organisers were anticipating. ‘We thought it was important that the participants have a real interest. One wonderful outcome of the School would be real research that participants can strengthen by using the knowledge, tools, and methods that we gave them during the Summer School,’ he adds.

Searching for a Nobel Prize

HSE and SDA are working with LHCb on another project that combines physics and computer science. The three are the organisers of a competition on Kaggle, an online platform for data prediction competitions. The 2014 competition had participants who searched for the Higgs boson, and an undergrad in the Faculty of Computer Science, Stanislav Semenov, took 7th place at the competition.

The current contest is devoted to a new type of physics. Participants will have to work on lepton flavour violations, and they will be handling both real data acquired from the LHCb experiment, as well as data from a simulator. The overall goal of the competition is to use collective efforts to find new methods for processing data that will help find the collapse of the tauonic leptons by three muon. No one has yet seen this collapse, but if found, it will be a find worthy of a Nobel Prize.

‘You come to the Kaggle platform, sign up, select the Flavours of Physics competition, familiarize yourself with the rules, and you get access to the data,’ Ustyuzhanin explains. ‘There’s a starter kit that helps beginners unravel the data, metrics, and format for sending their solution. It’s important that a person know how to program, as he will need to enjoy handling data. It’s also important for him to be interested in physics, particularly physics research. We have tried to make it so that participants don’t have to fully understand the intricacies of the experiment itself,’ he adds. The results of the best work might be used in the analysis of new data from LHCb in 2015.

Why Physics Is Becoming Popular

Everyone is interested in understanding the world around them, Andrey Ustyuzhanin is certain. For many, science is becoming somewhat of a hobby, and such an ‘opportune’ academic endeavour might be entirely productive. ‘A new viewpoint always has something to add, and the more new things there are, the higher are the chances that it will yield something useful,’ he says. ‘We see this in the Flavours of Physics competition, when people try to go beyond the limits we gave them. And the ways that exist for obviating these limits – even if they themselves are useless in physics – cause a chain reaction. Other participants have realized, aha you can’t do this or that, but maybe if you change your approach a little bit, something useful will come of it,’ Ustyuzhanin notes.

Astrophysicists have always been willing to work with those interested in the field. High-energy physics has always been closed off from those who are not professionals due to the specific culture and intricacies of working with data that an ‘outsider’ could not likely make sense of. But the situation is changing. ‘There’s a scientific movement underway, at least for certain experiments that are ready to share a portion of their data in order to attract outside expertise. Both sides benefit from this. CERN, for example, has released data from several 2010 experiments and is planning to publish data from the coming years,’ Ustyuzhanin comments.

The Smartphone as a New Weapon for the Physics Lover

CERN experiments are not the only projects that faculty from the Laboratory for Methods of Big Data Analysis (LAMBDA), which is headed by Ustyuzhanin, are working on. Crayfis in one example of a project that is opening up science to those interested in the field.

Crayfis is an international consortium made up of partners that include HSE and some of the largest American universities. The project searches for high-energy cosmic particles that end up in the Earth’s atmosphere. When they collide with the upper atmosphere, a sort of chain reaction begins, dispersing other particles. This brings about a particle ‘downpour’ that covers an area of about 1 square kilometre. The source of these particles remains one of the unsolved mysteries of our time.

These particles are very rare. According to some estimates, around 1 particle per 100 square meters falls to Earth each year. No one knows where they come from, but there are special observatories in South America that look for these particles, and it is clear that maintaining such labs and equipment costs a lot of money.

Crayfis is a project to search for these very particles, but with the help of mobile phones. A smartphone has sensitive enough features that its camera is able to find the products of such a particle’s decay.

‘We are developing the technical part. It is necessary to somehow filter camera images to lower noise and build a system for amassing data. This system would need to store information from all who participate in the project,’ Ustyuzhanin says. But a massive network of specially configured smartphones is needed to ‘catch’ a particle. The project will prove successful if it can get around a million participants worldwide. Anyone who uses their smartphones to participate in the project will have access to the data the experiment collects.

Science or Business?

‘The majority of people are more interested in using their abilities on things that make money, and in this sense, mathematicians and programmers are probably drawn more towards hedge funds or exchanges where you can implement an algorithm and instantaneously see how it reflects on a company’s finances. But there are people who are interested in more fundamental questions. In the same way that hackers study the specificities of other peoples’ programmes and systems, these people try to find the key to how the Universe works,’ Ustyuzhanin posits.

‘We view students in the laboratory as potential researchers who might become part of valuable physics experiments in the future,’ he continues. But the field for applying the knowledge attained in the lab is much broader. Being able to work with data allows you to switch between various scientific fields – from finance to medicine to particle physics.

The main thing is having interest and being ready to independently use the tools before you instead of waiting for someone else to do it for you, Ustyuzhanin says. ‘The ecosystem we have been able to create around us helps us bring in precisely these kinds of colleagues, and if any of the people reading this are like this, we would be happy to speak with you and see how we can help one another,’ he adds.

And elementary and high-school students interested in computer science or ways of using computer science in physics should be advised to visit HSE on September 13 for Programmers’ Day. Staff from LAMBDA and various IT companies will be there to explain how what they do is not only interesting, but also good for mankind.