Лекції

Четвер, 11 квітня

Vitalii Tymchyshyn 

Ph.D. (Theoretical Physics),M.Sc. (Biochemistry).
Researcher at Data Science Center, Kyiv Academic University.
Kyiv, Ukraine.

Topological Data Analysis for Cosmology

The topological data analysis of the Cosmic web is a rapidly developing field of study. Persistent homology - an ultimate topological tool to characterize a point cloud - is one of the most important and interesting advancements in this direction. In this lecture, we will marvel at the beauty of both the Cosmos and mathematical abstractions, while at the end everything will condense into a Python code. 

youtube recording link 

Час: 13:00 - 15:20 (звідси і надалі - GMT+3, Київ)

Olga Gogota

Ph.D. (Physics of the Nucleus, Elementary Particles and High Energies)
Researcher at Taras Shevchenko National University of Kyiv.
Kyiv, Ukraine.

Machine Learning for Neutrino Physics Experiments

Machine learning methods are used at different levels of the experiment: optimization of the experimental design, detector operation and data collection, physical modeling, data reconstruction, and physical inference. We will discuss recent trends in the application of machine learning tools in neutrino experiments.

youtube recording link 

Час: 15:30 - 16:40 

Mykhailo Koreshkov

MSc (Mathematics) - ongoing, KAU, Institute of mathematics of NASU;BSc (Applied Mathematics), Institute of physics and technology, NTUU "KPI".
Researcher/bioinformatician at Sysbio laboratory of the Institute of microbiology and genetics; software developer.Kyiv, Ukraine.

Approaches to Texture Classification for Biological Microscopic Photography

Last 6 months I have been working on my first professional research project at the Laboratory of systems biology, a part of the Institute of molecular biology and genetics. We studied some specific approaches to texture classification for biological microscopic photography.In this short talk I will show you what our typical texture classification pipeline looks like, introduce you to some popular image feature transforms that work well for this problem, and I will be happy to present some results of our research. Along the way, I will also share my experience of working in a research laboratory from a grad student perspective. See you soon!

youtube recording link

Час: 16:50-17:20

П'ятниця, 12 квітня

Olexandr Isayev

Ph.D. (Theoretical Chemistry). 
Associate Professor at Carnegie Mellon University.
Pittsburgh, USA.

Neural Networks Learning Quantum Mechanics

Deep learning is revolutionizing many areas of science and technology, particularly in natural language processing, speech recognition and computer vision. In this talk, we will provide an overview into latest developments of machine learning and AI methods and application to the problem of drug discovery and molecular design at Isayev’s Lab at CMU. We identify several areas where existing methods have the potential to accelerate computational chemistry research and disrupt more traditional approaches. We will present a deep learning model that approximate solution of Schrodinger equation. Focusing on parametrization for drug-like organic molecules and proteins, we have developed a single ‘universal’ model which is highly accurate compared to reference quantum mechanical calculations at speeds 10^6 faster.

youtube recording link

Час: 12:30 - 14:15

Tong Li

Ph.D. (Cell Biology).
Senior Software Developer at Bayraktar Lab (Cellular Genetics), Sanger Institute, Wellcome Genome Campus, Hinxton, Cambridge, UK.

FAIR Image Analysis: Unveiling Insights Through Reproducibility in Image-Based Spatial Transcriptomics

As I build the image analysis team at Sanger, the emphasis is on applying the Findable, Accessible, Interoperable, and Reusable (FAIR) principles to various image analysis techniques to tehcnologies such as spatial imaging, high-content screening and lightsheet imaging. Ensuring the findability, accessibility, interoperability, and reusability of such data analysis becomes paramount in this diverse technological landscape. It's crucial to note that reproducibility is not just about data but extends to metadata. During the presentation, I will showcase the Spatial Transcriptomics analysis pipeline as an illustrative example of how FAIR tools can remarkably enhance data analysis efficiency and promote broader scientific contributions through reusability and comprehensive reproducibility.
youtube recording link

Час: 14:20 - 15:50 

Jiří Vyskočil

Ph.D.
Scientific Computing CoreCenter for Advanced System Understanding, Helmholtz-Zentrum Dresden-Rossendorf.
Görlitz, Germany.

High-Energy Laser-Plasma Interactions

Recent advances in laser technology have opened many diverse fields of experiments where high-energy laser pulses interacting with plasmas generate streams of accelerated electrons or ions, produce gamma rays or electron-positron pairs, or emulate conditions found inside heavy planets or stars in a laboratory setting. Traditional simulation methods like the Particle-in-Cell (PIC) method, introduced in 1959, allow us to simulate the complex behavior of plasmas in such experiments. Getting past the proven path, efforts to enhance PIC codes through AI/ML are paving the way for further advancements in understanding and leveraging these high-energy interactions.
youtube recording link 

Час: 16:00 - 17:15 

Denys Klekots

M.Sc. (High Energy Physics) - ongoing.
Engineer of the nuclear physics laboratory at Taras Shevchenko National Unversity of Kyiv.Kyiv, Ukraine.

Machine Learning for Signal/Background Discrimination in High Energy Physics

Studies of fundamental matter properties and ground-level physics, like those conducted at the Large Hadron Collider, generate vast amounts of data, much of which is background noise. To measure rare physics events, it's essential to filter out this background noise and isolate signal events. Background suppression has always been a key aspect of high-energy physics data analysis. Historically, simple cuts on track properties were used, but nowadays, machine learning techniques are widely employed.
I'd like to present an overview of the signal event separation for high energy physics. I'm currently doing analysis of the LHCb collaboration data as part of my master's thesis. We'll focus on how we select training data for boosted decision trees and the validation and cross-checking procedures for our model on example of my master thesis work.
youtube recording link 

Час: 17:20 - 17:50 

Субота, 13 квітня

Oleksii Ignatenko

Doctor of Sciences.
Professor at Ukrainian Catholic University.
Lviv, Ukraine.

Shapley Values for Explainable Machine Learning

As machine learning models become increasingly complex and ubiquitous, the need for explainability and interpretability has emerged as a critical challenge. Explainable Machine Learning aims to bridge the gap between the opaque decisions made by these models and the human understanding of their reasoning processes. One powerful tool in the arsenal is the concept of Shapley values, derived from cooperative game theory.
This lecture will be about applying Shapley values in the context of explainable machine learning. Shapley values provide a principled approach to quantifying the contribution of each feature to a model's prediction, shedding light on the decision-making process. By breaking down the model's output into interpretable components, Shapley values enable us to understand the relative importance of different features and their interactions and consider practical examples.
youtube recording link 

Час: 11:00 - 12:15 

Mykola Korablov 

M.Sc. (Computer Science) - ongoing, KAU;B.Sc. (System Analysis), IASA,
NTUU "KPI"
Engineer-programmer,Department of Nonsmooth Optimization Methods,V.M. Glushkov Institute of Cybernetics of the NAS of Ukraine.
Kyiv, Ukraine.

Non-smooth Optimization in Machine Learning:
Review of N.Z. Shor's Methods and Their Applications 

Nowadays, numerical optimization is a principal core of training any learning algorithm: from linear regression to multi-layer neural networks. There are a number of challenges in the area, some of which may be solved by means of non-smooth optimization. In this lecture, we will review this area of optimization that was pioneered by Ukrainian mathematician Naum Zuselevych Shor and his team at V.M. Glushkov Institute of Cybernetics. We will cover main algorithms and crucial ideas they are based on. Moreover, we will provide a couple of examples how these methods can be applied to solving supervised learning problems. 

youtube recording link 

Час: 12:20 - 14:10 

Oleksiy Samoylenko

B.Sc. (Computer Science in Medicine and Biology).
Clinical Data Analyst at Mark company.Kyiv, Ukraine. 

Artificial Intelligence and Personalized Medicine

Modern artificial intelligence is saturating various fields of science, and one of the most promising areas is personalized medicine. This field, also known as precision or personalized medicine, combines the powerful capabilities of artificial intelligence with the healthcare industry. The result is a revolution in the way we think about diagnosis, treatment and patient care.

Буде викладено як запис в YouTube.

Oleksandr Petrenko 

MD (General Medicine).
Research Fellow at Medical University of Vienna & Predoctoral Fellow at CeMM Research Center for Molecular Medicine of the Austrian Academy of Sciences.
Vienna,  Austria.

Machine Learning in High-Throughput Biology Experiments: State of the Art

This talk will provide a comprehensive overview of the current state of machine learning (ML) applications in high-throughput biology experiments. It will focus on the challenges and opportunities in analyzing vast datasets generated by next-generation sequencing, such as bulk and single-cell RNA-sequencing, drug screens, and other assays. Emphasizing recent advancements in bioinformatics and computational biology, the talk will explore how ML algorithms are used to decipher complex biological data, uncover patterns in gene expression, protein interactions, and metabolic pathways.

youtube recording link 

Час: 14:50 - 15:20 

Development of Machine Learning Models for Biomedical Applications

This talk will explore the intersection of machine learning (ML) and biomedical sciences, highlighting the development and application of ML models for clinical data. It will be discussed how these models can enhance diagnostic accuracy, predict patient outcomes, and facilitate personalized medicine. 

youtube recording link 

Час: 16:30 - 17:00 

Nitesh V. Chawla

Ph.D.
Professor at the University of Notre Dame.Notre Dame, USA.

AI for Science & Society

This presentation will focus on the pivotal role of computing, including AI, in driving significant advancements, fostering multidisciplinary collaborations, and creating opportunities for scientific and societal impact. I will showcase our contributions to machine learning, with a special focus on graph-learning and how they are directed for both science and society.

youtube recording link 

Час: 15:30 - 16:25 

Alina Frolova

Ph.D. 
Institute of Molecular Biology and Genetics of NASU,
Researcher at Data Science Center.
Kyiv, Ukraine.

Metagenomics Applications

Studying microorganisms and their impact on humans can be a complicated task. Not only the inability to cultivate a majority of species but also the lack of reference genomes that can be used for comparative and evolutionary studies contribute to the complexity of the problem. Metagenomics has become a powerful approach that allows for studying microbial communities by collecting samples directly from the environment of interest, thus avoiding the need to rely on genome annotations. Moreover, the ever-increasing volumes of metagenomic data led to the advent of a variety of algorithmic approaches designed to operate on raw unannotated sequences, making up a significant portion of bioinformatics methods. In the talk, we will delve into the most notable applications of metagenomics, such as antimicrobial-resistance study and environmental monitoring, and how humans can benefit from it.

youtube recording link 

Час: 17:00 - 19:15