IDAO 2022 Challenges Participants to Work with Graph Neural Networks

© Idao.world

On February 1, the IDAO 2022 was launched. The HSE Faculty of Computer Sciences, Yandex and Otkritie Bank (platinum partner) are holding it for the fifth time. The first round of the olympiad will be held in an online format on Yandex.Contest - an online platform for organizing programming competitions.

Although the Olympiad has officially started, registration is open until February 13.

On February 3, the task for the first round was presented, which is devoted to predicting the properties of two-dimensional crystals of various configurations. Opening the event was Konstantin Novoselov, Professor at the National University of Singapore and winner of the 2010 Nobel Prize in Physics. He delivered a lecture on two-dimensional crystals and their properties.

Konstantin Novoselov

‘The synthesis of two-dimensional crystals with specified properties is a fundamentally new area of interdisciplinary research. One way to create new crystals is to obtain alloys and the controlled introduction of defects. Structural defects significantly affect physical and chemical properties, such as conductivity, photoluminescence, and magnetism. The space of possible configurations is huge, so the use of machine intelligence methods will significantly reduce the time to search for new key configurations with the specified properties.’

Abdalaziz Rashid Al-Maeeni, Research Assistant at the LAMBDA laboratory, doctoral student at the HSE Faculty of Computer Sciences, and one of the task developers, also spoke at the qualifying round presentation. He covered the task development and what will be useful for its solution.

Abdalaziz Rashid Al-Maeeni

‘After the discovery of graphene, interest in two-dimensional materials has grown in various industries due to their interesting electrical, optical, magnetic and other properties. Simulation of the density functional theory, which takes a very long time even on the most powerful supercomputers, is used to calculate the physical characteristics and properties of such materials.

The cooperation programme involving the National University of Singapore, HSE LAMBDA Laboratory and Innopolis University aims to use machine learning to accelerate the assessment of material properties. To get started, two-dimensional transition metal dichalcogenides seem promising.

Regarding the task itself, it consists in predicting the width of the forbidden energy zone of two-dimensional materials with point defects, which is directly related to the electrical properties of materials. The defects in the crystal change the width of this zone, so it is very important to understand how different types of defects affect it. The data is obtained as a result of numerical simulation. HSE LAMBDA Laboratory, researchers from the National University of Singapore and Innopolis University are working together on this project.

To solve the problem, participants will need to apply knowledge in machine learning, especially graph neural networks, along with other architectures that have necessary inductive biases and physical symmetries. Basic knowledge in physics, mainly materials science, will also be an advantage.

Your successful contribution can lead to breakthroughs in catalysis, drug delivery, biomedical imaging, biosensors, optoelectronic devices, photothermal therapy and many other fields. Depending on the level of solutions presented, we can even help combine the best works into a joint publication.’

Watch IDAO 2022 Qualifying Round Task Presentation