Leonid Legashev

Mr. Leonid Legashev

orcid

Research Focus & Metrics

Academic Career Length

13Years (2014-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
15 October 2024 in Toxics
Verified
Journal Article
Open access
PDF file
IKIvan Khokhlov
Ivan Khokhlov
Leonid LegashevLeonid Legashev
Leonid Legashev
IBIrina Bolodurina
Irina Bolodurina
ASAlexander Shukhman
Alexander Shukhman
DSDaniil Evgenievich Shoshin
Daniil Evgenievich Shoshin
SKSvetlana Valentinovna Kolesnik
Svetlana Valentinovna Kolesnik
Predicting the toxicity of nanoparticles plays an important role in biomedical nanotechnologies, in particular in the creation of new drugs. Safety analysis of nanoparticles can identify potentially harmful effects on living organisms and the environment. Advanced machine learning models are used to predict the toxicity of nanoparticles in a nutrient solution. In this article, we performed a comparative analysis of the current state of research in the field of nanoparticle toxicity analysis using machine learning methods; we trained a regression model for predicting the quantitative toxicity of nanoparticles depending on their concentration in the nutrient solution at a fixed point in time with the achieved metrics values of MSE = 2.19 and RMSE = 1.48; we trained a multi-class classification model for predicting the toxicity class of nanoparticles depending on their concentration in the nutrient solution at a fixed point in time with the achieved metrics values of Accuracy = 0.9756, Recall = 0.9623, F1-Score = 0.9640, and Log Loss = 0.1855. As a result of the analysis, we concluded the good predictive ability of the trained models. The optimal dosages for the nanoparticles under study were determined as follows: ZnO = 9.5 × 10−5 mg/mL; Fe3O4 = 0.1 mg/mL; SiO2 = 1 mg/mL. The most significant features of predictive models are the diameter of the nanoparticle and the nanoparticle concentration in the nutrient solution.

1 September 2019
Verified
Conference Paper
Closed access
NBNadezhda Bakhareva
Nadezhda Bakhareva
ASAlexander Shukhman
Alexander Shukhman
AMArtem Matveev
Artem Matveev
PPPetr Polezhaev
Petr Polezhaev
YUYury Ushakov
Yury Ushakov
Leonid LegashevLeonid Legashev
Leonid Legashev
This paper proposes algorithms for detecting attacks in enterprise networks based on the analysis of network traffic. The CICIDS2017 dataset was used to compare machine learning methods for binary classifying (attack or regular traffic), as well as for multiclass classification to identify the classes of typical attacks such as DoS/DDoS, PortScan, BruteForce, WebAttack, Bot and Infiltration. The balanced accuracy score is used as the main metric for assessing the accuracy of classification. The main advantage of this metric in adequately estimating the accuracy of classification algorithms, considering the strong imbalance in the number of labeled records for each class of dataset. As a result of the experiment, it was found that the CatBoost and LightGBM algorithms work well for both binary classification and multiclass classification of malicious traffic into several attack groups.

1 January 2019 in Procedia Computer Science
Verified
Journal Article
Open access
Leonid LegashevLeonid Legashev
Leonid Legashev
TLT.V. Letuta
T.V. Letuta
PPP.N. Polezhaev
P.N. Polezhaev
ASA.E. Shukhman
A.E. Shukhman
YUYu.A. Ushakov
Yu.A. Ushakov

23 April 2025 in Applied Sciences
Verified
Journal Article
Open access
PDF file
Leonid LegashevLeonid Legashev
Leonid Legashev
ASAlexander Shukhman
Alexander Shukhman
VBVadim Badikov
Vadim Badikov
VKVladislav Kurynov
Vladislav Kurynov
In the development of goal-oriented dialogue systems, neural network topic modeling and clustering methods are traditionally used to extract user intentions and operator response scenario blocks. The emergence of generative large language models allows one to radically change the approach to generate dialogue scenarios in the form of a graph with context preservation. In this article we analyzed seven popular large language models on prepared test prompts for Russian and English languages for intent mining and named entity recognition. The present study aimed to investigate the effectiveness of two methods for constructing dialogues in goal-oriented dialogue systems: the heuristic-based approach with additional training on labeled data and the prompt-based approach without such training. The primary objective was to evaluate the impact of incorporating labeled dialogue data on the quality of constructed dialogues, with a focus on dialogue context. The study emphasized the need for dialogue systems to consider the dialogue context in constructing goal-oriented dialogues. The two approaches were compared for the MultiWOZ 2.2 and MANTiS dialogue corpora on a locally deployed LLaMA model. The results showed that the LlaMA model without training on labeled dialogues achieved a BERTScore metric value of 0.75 for the MultiWOZ dataset and 0.72 for the MANTiS dataset, and the LlaMA model with training on labeled dialogues achieved a BERTScore metric value of 0.85 for the MultiWOZ dataset and 0.82 for the MANTiS dataset. This finding has practical implications for the development of more effective dialogue systems in the field of customer service that can engage users in more productive and meaningful machine-to-human interactions.

1 April 2018
Verified
Conference Paper
Closed access
ASAlexander E. Shukhman
Alexander E. Shukhman
IBIrina P. Bolodurina
Irina P. Bolodurina
PPPetr N. Polezhaev
Petr N. Polezhaev
YUYury A. Ushakov
Yury A. Ushakov
Leonid LegashevLeonid Legashev
Leonid Legashev
This paper describes the project of an automated adaptive system for individual support of talented students with the educational IT infrastructure. A model of the process of self-learning and self-development of talented students in the educational IT infrastructure is developed on the basis of a multilevel competence system using Petri nets. To design optimal student's individual learning paths algorithms of the ant colony have been used. Methods for calculating the relative values of the student's rating are proposed based on accounting their academic achievements in a specific subject area. A preliminary version of the Web portal has been developed to provide the storage of talented student's profiles with their achievements and ratings, information on courses and training competitions, ILP planning, user interaction, integration with external educational resources.

8 September 2015
Verified
Conference Paper
Open access
ASAlexander Shukhman
Alexander Shukhman
PPPetr Polezhaev
Petr Polezhaev
YUYuri Ushakov
Yuri Ushakov
Leonid LegashevLeonid Legashev
Leonid Legashev
VTVeniamin Tarasov
Veniamin Tarasov
NBNadezhda Bakhareva
Nadezhda Bakhareva
This paper presents the architecture of information security systems for enterprise software-defined networks. Its main advantages are modularity, expandability, unified network configuration and monitoring, flexibility, load scalability, efficient hardware use, total check of network traffic. The architecture has been implemented as a firewall. Our firewall algorithm was tested by the Mininet simulator. The results of network performance evaluation have confirmed the firewall efficiency by delays and bandwidth.