Seminars

The Laboratory conducts international and periodic seminars, organizes public lectures and participates in various HSE events.

Information about planned events is published on the announcements page of the laboratory. This page contains the events of the current year. You can find out about the events of previous years on the Events Archive page.

Seminars are open access for everyone.
Admission is free for students, postgraduates, teachers and staff of the Higher School of Economics.
If you need to order a pass to the HSE building, please inform us by e-mail: lantropova@hse.ru 

On April 15, 2026, at 12:00 PM, as part of the "Mathematical Models of Information Technologies" seminar at the International Laboratory of Intelligent Systems and Structural Analysis, the Center for Language and Semantic Technologies, and the Department of Data Analysis and Artificial Intelligence, in collaboration with the Department of Software Engineering, PhD Sergey Aleksandrovich Stupnikov (Federal Research Center "IUC" of the Russian Academy of Sciences) will present a report on "Methods and Tools for Verifiable Integration of Heterogeneous Data" in an online conference format.

Abstract:

Semantic data integration is necessary to ensure uniform access, interpretation, and subsequent use of heterogeneous data sources: their volume, diversity, and complexity are constantly growing. Over the past twenty years, numerous technologies and systems have been developed to support semantic data integration. However, data integration verification—that is, formally proving that the semantics of the data and operations on it are preserved during data integration—remains a complex task.

This paper presents methods for verifying data integration at the levels of data models, data schemas, and the data itself. The methods are based on defining the semantics of integration programs in a formal specification language supported by automated proof tools. These methods are intended for use in developing systems for integrating heterogeneous data, which are necessary for solving problems in research infrastructures across various subject areas and in industrial information systems.

On February 4, 2026, at 11:00 AM, as part of the "Mathematical Models of Information Technologies" seminar at the International Laboratory of Intelligent Systems and Structural Analysis, Center for Language and Semantic Technologies, Department of Data Analysis and Artificial Intelligence, Gleb Yuryevich Kuzmin, Junior Researcher, Department No. 74, FRC ICS RAS, and AIRI Researcher, will present a report titled "Biaslessness and Uncertainty Estimates in Text Analysis Problems" in an online conference format.

The seminar will address issues of improving the security of natural language processing models through the combined use of uncertainty estimation and bias removal methods. It is shown that bias removal methods can significantly degrade the quality of uncertainty estimates, which is critical for tasks with a high error cost. To address this issue, methods for effectively combining the above methods are proposed. A proposed approach for selective bias removal is also discussed, which improves the unbiasedness of model predictions by using uncertainty estimates through selective bias removal for the most biased predictions. The effectiveness of the proposed methods has been experimentally confirmed on a number of natural language processing tasks.

On January 28, 2026, at 12:00 PM, Tasnima Ravilevna Sadekova (Huawei) will present a report on "Studying Models for Voice Cloning and Improving Their Performance" as part of the "Mathematical Models of Information Technologies" workshop at the International Laboratory of Intelligent Systems and Structural Analysis, Center for Language and Semantic Technologies, and Department of Data Analysis and Artificial Intelligence. The workshop will discuss voice cloning problems and key approaches to solving them. Particular attention will be paid to the new diffusion architecture in comparison with other models. Furthermore, improvements to key characteristics of such systems, such as generation speed and control over the emotionality of speech, will be discussed.

On December 29, 2025, the annual international seminar "LOGIC MATTERS 2025" was held by the International Laboratory of Intelligent Systems and Structural Analysis of the Center for Language and Semantic Technologies, Department of Data Analysis and Artificial Intelligence, Faculty of Computer Science. The international seminar "Logic Matters" aims to discuss scientific issues in mathematics, machine learning, formal logic, digital linguistics, and artificial intelligence. Registration for the event is available at https://cs.hse.ru/ai/clst/issa/logmatters/

On November 26, 2025, at 1:00 PM, as part of the joint seminar "Mathematical Models of Information Technologies," the Center for Language and Semantic Technologies, Department of Data Analysis and Artificial Intelligence, and the International Laboratory of Intelligent Systems and Structural Analysis, led by S.O. Savva Viktorovich Ignatyev will present a talk on "Implicit Neural Representations for 3D Generation and 3D Reconstruction from Multiple Viewpoints" at the Kuznetsov Moscow State University in Moscow, Russia, in an online conference format. The seminar will present the results of a dissertation on implicit neural representations (INRs) and describe their current shortcomings. To overcome these shortcomings, the proposed work develops and describes a number of methods:

training a hypernetwork INR as a generator for a generative adversarial network (GAN);
obtaining aligned 3D models parameterized by a single INR from a given set of text descriptions;
an algorithm for fast rendering and reconstruction of an implicit surface;
an approach for augmenting a partially observed surface;
an unsupervised GAN-based image generation method for separating shape and appearance in images, in which deformation and texture maps are generated by separate INR generators.

On November 7, 2025, at 12:00 PM, Ivan Aleksandrovich Karpushin will present a report on "Models and Algorithms for Training Stochastic Neural Networks for Extracting High-Precision Representations in Pattern Recognition Problems" as part of the joint seminar "Mathematical Models of Information Technologies" held by the Center for Language and Semantic Technologies, the Department of Data Analysis and Artificial Intelligence, and the Laboratory of Intelligent Systems and Structural Analysis, led by S.O. Kuznetsov. The seminar will present the results of his dissertation, which focuses on the development of models and algorithms for training stochastic neural networks. The research aims to develop methods for solving classification and training problems in metric spaces.

On October 10, 2025, at 12:00 PM, as part of the joint seminar "Mathematical Models of Information Technologies" held by the Center for Scientific Research, the Department of Data Analysis and Artificial Intelligence, and the Laboratory of Intelligent Systems and Structural Analysis, led by S.O. Kuznetsov, Dmitry Olegovich Brykin, a postgraduate student in the Joint Department of the MIPT Computer Information Systems (CIS), will present a dissertation for a candidate of science degree, entitled "Research of Algorithms for Processing and Statistical Analysis of Data and the Development of a Mechanism for Improving the Accuracy of Sales Forecasting in Corporate Information Systems." The topic will be presented in an online conference format.
Modern corporate information systems are becoming a vital source of data for analysis and forecasting. Thanks to the development of technologies for storing and processing large amounts of information, as well as the implementation of data mining methods, such systems are no longer limited to simply automating accounting; they are becoming powerful decision support tools. Among the many similar systems used in businesses worldwide, solutions based on the SAP and Microsoft Dynamics platforms occupy a special place. In Russia and the CIS, 1C:Enterprise has become the most popular system, occupying a leading position in domestic software for automating management accounting and business processes. However, implementing forecasting algorithms in the 1C environment poses a number of non-obvious challenges related to a number of technical features of the platform: the lack of native support for complex matrix operations, limited computing capabilities, including in cloud solutions, and the specific nature of the 1C interpreted language. These features create significant obstacles to the implementation of modern, computationally intensive forecasting algorithms. Under these conditions, it is necessary to develop and adapt models and implement algorithms capable of ensuring high forecast accuracy with limited computing resources and the specific capabilities of the platform.

On September 26, 2025, at 12:30 PM, as part of the joint seminar "Mathematical Models of Information Technologies" of the Central Research Institute of Standardization, the Department of Data Analysis and Artificial Intelligence, and the Laboratory of Intelligent Systems and Structural Analysis, led by S.O. Alexey Sergeevich Vatolin will present a report on "Constructing Models for Vectorizing Scientific Texts and Creating an Open Infrastructure for Evaluating Them" at the Kuznetsov University in an online conference format.
The report will focus on constructing models for vectorizing scientific texts and creating an open infrastructure for evaluating them.

On September 22, 2025, at 3:30 PM, as part of the joint seminar "Mathematical Models of Information Technologies" held by the Center for Scientific Research, the Department of Data Analysis and Artificial Intelligence, and the Laboratory of Intelligent Systems and Structural Analysis, led by S.O. Kuznetsov, Nikita Denisovich Shaimov, a postgraduate student at the National Research University Higher School of Economics and a research intern at the National University of Applied Mathematics and Information Systems, will present a report titled "Methods and Algorithms for Improving the Representation of Cyclic Behavior in Process Models Synthesized from Event Logs." The report is presented as an online conference.

Abstract:
This report focuses on methods for constructing and visualizing process models based on event logs, taking into account the correct representation of cyclic behavior. Correctly representing cyclic behavior is an important task, as its distortion reduces the accuracy of models and the quality of their subsequent analysis. The paper presents two algorithms for detecting direct sequence graphs that maintain full correspondence with the event log and prevent the formation of cycles not supported by the data. A visualization method combining DFG and Sankey diagrams is also proposed, allowing for the visual identification of case groups and analysis of their trajectories. Experimental studies on real and synthetic journals confirm the effectiveness of the proposed methods and their practical applicability.

On March 26, 2025, at 3:00 PM, as part of a joint seminar of the Center for Language and Semantic Technologies of the ISSA and DADII, Elena Viktorovna Chistova, Junior Research Fellow at the Federal Research Center for Information Systems of the Russian Academy of Sciences, will present a report on the topic: "Methods for Analyzing the Rhetorical Structure of Russian-Language Texts."

Presenter: Elena Viktorovna Chistova

Topic: "Methods for Analyzing the Rhetorical Structure of Russian-Language Texts."
Abstract: Many natural language processing tasks require text analysis beyond a single sentence—discourse analysis. One of the most widely used theories for describing the discursive structure of a text is rhetorical structure theory. The report will present methods for surface and full-text analysis of the rhetorical structure of Russian-language texts and their applications to classification, argumentation analysis, and coreference resolution problems, developed as part of a dissertation.

On March 19, 2025, at 3:00 PM, as part of the joint seminar "Mathematical Models of Information Technologies" held by the Department of Data Analysis and Artificial Intelligence and the Laboratory of Intelligent Systems and Structural Analysis, led by S.O. Kuznetsov, Mikhail Mikhailovich Tikhomirov, PhD, research fellow at the Moscow State University Research Computing Center, will present a report on "Adaptation of Large Language Models for the Russian Language" in an online conference format.

Abstract:
Despite the rapid development of large language models, particularly in the area of ​​multilingualism, the quality and effectiveness of such LLMs in Russian remains inferior to those in English. There are two main ways to solve these problems: 1. teaching LLM from scratch, which requires millions of dollars, or 2. adapting existing models into Russian.

Date: 11.05.2022
Topic: "Informative discourse feature selection for analysis of textual data". 
Speaker: Elizaveta Goncharova, International Laboratory for Intelligent Systems and Structural Analysis.
Abstract: The presented research is dedicated to the analysis of the modern pre-trained language models (LMs) and their ability to inject linguistic features, such as discourse, during solving natural language processing tasks.Recent pre-trained LMs have shown state-of-the-art results on a bunch of NLP tasks, however, these models still suffer from the insufficient linguistic representation of a text that eventually leads to a low level of language understanding. In order to improve these models’ performance novel methods of discourse structure encoding have been proposed in the research. The introduced approaches allow us to incorporate discourse features into the LMs explicitly during pre-training or fine-tuning procedures without requiring significant modifications to the model's architecture. We provided the experimental evaluation of the discourse-aware models on various complex NLP tasks which are argumentation classification (AC), question answering (QA), and text summarization, and concluded that the modified models achieve results as good as or better than other discourse-free or more complex discourse-aware models on the well-known NLP benchmarks. Finally, the influence of discourse features on the models' explainability is considered. In the research, we introduced an independent explainability pipeline that is able to reveal relevant text spans based on the discourse relations assigned to them that can be used to explain deep learning models' decisions in supervised NLP tasks.

Date: 15.03.2022
Topic: "Less is more" based heuristics for Minimum sum-of-squares clustering".
Speaker: Nenad Mladenovich, Professor of the Faculty of Industrial Systems Design, Khalifa University, Abu Dhabi, UAE.
Abstract: The "Less is more" (LIMA) approach to optimization has recently been proposed. Its main idea is to include simplicity, in addition to efficiency and accuracy, in the comparison of the two algorithms. In this report, in addition to classical local searches for solving the problem of unsupervised learning with a minimum sum of squares, i.e. k-means, h-means and j-means, I will also present some new simple heuristics for a huge dataset with several million objects: (i) One-pass k-means and (ii) decomposition/aggregation of k-means. The results of the calculations will show the advantages of including the simplicity criterion in big data analysis.