Персона:
Сбоев, Александр Георгиевич

Организационные подразделения

Организационная единица

Цель ИЯФиТ и стратегия развития - создание и развитие научно-образовательного центра мирового уровня в области ядерной физики и технологий, радиационного материаловедения, физики элементарных частиц, астрофизики и космофизики.

Фамилия

Сбоев

Имя

Александр Георгиевич

Полная страница элемента

Результаты поиска

Теперь показываю 1 - 10 из 62

Только метаданные
Evaluation of Machine Learning Methods for Relation Extraction Between Drug Adverse Effects and Medications in Russian Texts of Internet User Reviews
(2022) Selivanov, A.; Rybka, R.; Moloshnikov, I.; Rylkov, G.; Sboev, A.; Сбоев, Александр Георгиевич
© Copyright owned by the author(s) under the terms of the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License (CC BY-NC-ND 4.0).The research considers an automatic extraction of relations between mentions of medications and adverse drug reactions in Russian-language drug reviews. This text analyzing method might be useful for pharmacovigilance and medicines reprofiling. Its application to Russian-language reviews hasn’t been studied yet due to the lack of corpora with relation annotation in Russian. The study is aimed at solving this problem. It is based on the original dataset gathered by our group. It consists of annotated relations between entities from the Russian Drug Review Corpus, that contains the Internet users’ reviews on medications in Russian language. Computational experiments were carried out on developed corpora using classical machine learning methods, as well as a more advanced neural network model based on Transformer layers – XLM-RoBERTa-sag. The list of applied classical machine learning methods consists of support vector machine, logistic regression, Naive Bayes classifier and gradient boosting. The concatenation of TF-IDF entity vectors of character n-grams was used as a text representation. Based on a set of experiments, the following hyperparameters of these methods were selected: the size of n-grams and the limitation on the frequency of occurrence of n-grams (too rare or too frequent n-grams were excluded from the feature vector). For XLM-RoBERTa-sag, the input data is represented as usual for such type of models (language models based on Transformer topology). The following input text representation types were considered during the experiments: a whole text, a text of target entity pairs; a text of target entity pairs with words between them; a text of target entity pairs and the whole input text, the latter input type is the one that maximizes accuracy. It is shown that XLM-RoBERTa-sag model achieves a result of 95%, according to the macro-averaged f1 metric, which is the state-of-the-art result of recognition of the relations between mentions of adverse drug reactions and medications in Russian-language online reviews. The Naive Bayes classifier with multivariate normal distribution achieves the best result among classical machine learning methods: 75%, which exceeds the result of random label generation by 21%.
Только метаданные
Adverse Drug Reaction Concept Normalization in Russian-Language Reviews of Internet Users
(2022) Sboev, A.; Rybka, R.; Gryaznov, A.; Moloshnikov, I.; Sboeva, S.; Rylkov, G.; Selivanov, A.; Сбоев, Александр Георгиевич
Mapping the pharmaceutically significant entities on natural language to standardized terms/concepts is a key task in the development of the systems for pharmacovigilance, marketing, and using drugs out of the application scope. This work estimates the accuracy of mapping adverse reaction mentions to the concepts from the Medical Dictionary of Regulatory Activity (MedDRA) in the case of adverse reactions extracted from the reviews on the use of pharmaceutical products by Russian-speaking Internet users (normalization task). The solution we propose is based on a neural network approach using two neural network models: the first one for encoding concepts, and the second one for encoding mentions. Both models are pre-trained language models, but the second one is additionally tuned for the normalization task using both the Russian Drug Reviews (RDRS) corpus and a set of open English-language corpora automatically translated into Russian. Additional tuning of the model during the proposed procedure increases the accuracy of mentions of adverse drug reactions by 3% on the RDRS corpus. The resulting accuracy for the adverse reaction mentions mapping to the preferred terms of MedDRA in RDRS is 70.9% F1-micro. The paper analyzes the factors that affect the accuracy of solving the task based on a comparison of the RDRS and the CSIRO Adverse Drug Event Corpus (CADEC) corpora. It is shown that the composition of the concepts of the MedDRA and the number of examples for each concept play a key role in the task solution. The proposed model shows a comparable accuracy of 87.5% F1-micro on a subsample of RDRS and CADEC datasets with the same set of MedDRA preferred terms.
Только метаданные
Application of Machine Learning to Construct Solitons of Generalized Nonlinear Schrodinger Equation
(2024) Sboev, A. G.; Kudryashov, N. A.; Moloshnikov, I. A.; Nifontov, D. R.; Rybka, R. B.; Сбоев, Александр Георгиевич; Кудряшов, Николай Алексеевич; Молошников, Иван Александрович; Нифонтов, Даниил Романович; Рыбка, Роман Борисович
Открытый доступ
ПРОГРАММА ПРЕОБРАЗОВАНИЯ ВЕСОВ НЕЙРОННОЙ СЕТИ К ЗАДАННЫМ РЕСУРСНЫМ ОГРАНИЧЕНИЯМ НЕЙРОМОРФНЫХ ВЫЧИСЛИТЕЛЬНЫХ УСТРОЙСТВ НА ОСНОВЕ КЛАСТЕРИЗАЦИИ
(НИЯУ МИФИ, 2022) Серенко, А. В.; Сбоев, А. Г.; Рыбка, Р. Б.; Сбоев, Александр Георгиевич
Программа преобразует веса нейронной сети так, чтобы они удовлетворяли ограничениям, типичным для цифровых нейроморфных устройств: получаемая в результате нейронная сеть имеет целочисленные веса, и входящие веса каждого нейрона могут принимать не более заданного количества уникальных абсолютных значений. Например, если заданное количество - 4, то каждый нейрон имеет 4 целочисленных параметра, один из которых - ноль, а каждый входящий вес нейрона может принимать 7 значений: одно из трёх ненулевых значений со знаком плюс, со знаком минус, или ноль. Получение таких весов из весов исходной обученной нейронной сети производится с помощью кластеризации и дообучения; в ходе дообучения на этапе вычисления активаций сети веса подвергаются кластеризации и округлению до целого, а изменению по градиенту функции ошибки подвергаются центры кластеров. Программа требует для своей работы установленных открытых программных пакетов tensorflow и tensorflow-model-optimization. Поддерживаются нейронные сети полносвязной и свёрточной топологии.
Только метаданные
A gender identification of text author in mixture of Russian multi-genre texts with distortions on base of data-driven approach using machine learning models
(2019) Gudovskikh, D.; Moloshnikov, I.; Rybka, R.; Sboev, A.; Сбоев, Александр Георгиевич
© 2019 Author(s).In this work we investigate a wide set of machine learning models of data-driven approaches (Long Short-Term Memory networks, Convolutional neural networks, multilayer perceptrons, Random Forest Classifiers, Logistic Regression and Gradient Boosting Classifiers with different sets of features) to identify the gender of author in Russian multi-genre texts in the case of existing style distortions and gender deceptions in training and testing sets. We consider and evaluate accuracy for the following situations: the influence of style distortions and gender deceptions in training texts for different genre, and the case when such deception is present only in test results. A comparison with known literature data is presented. The set of data corpora includes: one collected by a crowdsourcing platform, essays of Russian students (RusPersonality), Gender Imitation corpus, and the corpora used at Forum for Information Retrieval Evaluation 2017 (FIRE), containing texts from Facebook, Twitter and Reviews. We present the analysis of numerical experiments based on different features (morphological data, vector of character n-gram frequencies, LIWC and others) of input texts along with various machine learning models. The presented results, obtained on a wide set of data-driven models, establish the accuracy level for the task to identify gender of an author of a Russian text in the multi-genre case and analyzed the effect of the presence of deception in the test and training sets.
Открытый доступ
ПРОГРАММА НАСТРОЙКИ МОДЕЛИ ВЫДЕЛЕНИЯ СВЯЗАННЫХ ИМЕНОВАННЫХ СУЩНОСТЕЙ С ОЦЕНКОЙ ЕЁ ЭФФЕКТИВНОСТИ
(НИЯУ МИФИ, 2022) Молошников, И. А.; Сбоев, А. Г.; Селиванов, А. А.; Грязнов, А. В.; Рыльков, Г. В.; Рыбка, Р. Б.; Молошников, Иван Александрович; Сбоев, Александр Георгиевич; Рыбка, Роман Борисович
Программа предназначена для настройки модели выделения связанных именованных сущностей SpERT с расширением функциональности и включает в себя следующие процедуры: подготовка данных с использованием различных «токенизаторов», конвертация данных во внутренний формат системы SpERT из .json и обратно, векторное представление текстовых данных с использованием различных предварительно обученных языковых моделей, оценка модели выделения связанных именованных сущностей с учётом разрывной и пересекающейся разметки текстовых данных. В качестве входных данных программа использует файл .json, который представляет собой список, каждый объект в котором характеризует один входной текст. Такой объект содержит поля «text» (текст в исходном виде), entities (список выделенных в тексте сущностей), relations (список выделенных в тексте связей, используется для процедуры оценки модели). В качестве выходных данных программа представляет файл с предсказаннными связями между сущностями, отчёт по оценке эффективности модели.
Только метаданные
Deep Neural Networks Ensemble with Word Vector Representation Models to Resolve Coreference Resolution in Russian
(2020) Gryaznov, A.; Sboev, A.; Rybka, R.; Сбоев, Александр Георгиевич; Рыбка, Роман Борисович
© 2020, Springer Nature Switzerland AG.In this paper we present a novel neural networks ensemble to solve the task of coreference resolution in Russian texts. The ensemble consists of several neural networks, each based on recurrent Bidirectional long short-term memory layers (BiLSTM), attention mechanism, consistent scoring with selection of probable mentions and antecedents. The applied neural network topology has already shown state-of-the-art results in English for this task, and is now adapted for the Russian language. The resulting coreference markup is obtained by aggregating output scores from several blocks of independently trained neural network models. To represent an input source text, a combination of word vectors from two language models is used. We study the dependence of the coreference detection accuracy on various combinations of models of vector representation of words along with two tokenization approaches: gold markup or UDpipe tools. Finally, to show the improvement made by our ensemble approach, we present the results of experiments with both RuCor and AnCor datasets.
Только метаданные
Memristor-based spiking neural network with online reinforcement learning
(2023) Vlasov, D.; Minnekhanov, A.; Rybka, R.; Davydov, Y.; Sboev, A. G.; Сбоев, Александр Георгиевич
Neural networks implemented in memristor-based hardware can provide fast and efficient in-memory computation, but traditional learning methods such as error back-propagation are hardly feasible in it. Spiking neural networks (SNNs) are highly promising in this regard, as their weights can be changed locally in a self-organized manner without the demand for high-precision changes calculated with the use of information almost from the entire network. This problem is rather relevant for solving control tasks with neural-network reinforcement learning methods, as those are highly sensitive to any source of stochasticity in a model initialization, training, or decision-making procedure. This paper presents an online reinforcement learning algorithm in which the change of connection weights is carried out after processing each environment state during interaction-with-environment data generation. Another novel feature of the algorithm is that it is applied to SNNs with memristor-based STDP-like learning rules. The plasticity functions are obtained from real memristors based on poly-p-xylylene and CoFeB-LiNbO
Открытый доступ
ПРОГРАММА МОДЕЛИРОВАНИЯ ОБУЧЕНИЯ СПАЙКОВОЙ СЕТИ С STDP С КОРРЕЛЯЦИОННЫМ КОДИРОВАНИЕМ ВХОДНЫХ ДАННЫХ
(Федеральное государственное бюджетное учреждение «Национальный исследовательский центр «Курчатовский институт», 2022) Серенко, А. В.; Сбоев, А. Г.; Рыбка, Р. Б.; Сбоев, Александр Георгиевич; Рыбка, Роман Борисович
Программа реализует решение типовых классификационных задач ирисов Фишера и Висконсинского рака груди путём обучения спайковой нейронной сети с синаптической пластичностью Spike-Timing-Dependent Plasticity (STDP). Входные данные кодируются взаимной корреляцией входных спайковых последовательностей. Результат классификации декодируется путём сравнения корреляции выходных последовательностей спайков со входными на тестовой выборке с распределением этих корреляций на обучающей выборке.
Только метаданные
Spiking neural network with local plasticity and sparse connectivity for audio classification
(2024) Rybka, R. B.; Vlasov, D. S.; Manzhurov, A. I.; Sboev, A. G.; Рыбка, Роман Борисович; Сбоев, Александр Георгиевич
Purpose. Studying the possibility of implementing a data classification method based on a spiking neural network, which has a low number of connections and is trained based on local plasticity rules, such as Spike-Timing-Dependent Plasticity. Methods. As the basic architecture of a spiking neural network we use a network included an input layer and layers of excitatory and inhibitory spiking neurons (Leaky Integrate and Fire). Various options for organizing connections in the selected neural network are explored. We have proposed a method for organizing connectivity between layers of neurons, in which synaptic connections are formed with a certain probability, calculated on the basis of the spatial arrangement of neurons in the layers. In this case, a limited area of connectivity leads to a higher sparseness of connections in the overall network. We use frequency-based coding of data into spike trains, and logistic regression is used for decoding. Results. As a result, based on the proposed method of organizing connections, a set of spiking neural network architectures with different connectivity coefficients for different layers of the original network was implemented. A study of the resulting spiking network architectures was carried out using the Free Spoken Digits dataset, consisting of 3000 audio recordings corresponding to 10 classes of digits from 0 to 9. Conclusion. It is shown that the proposed method of organizing connections for the selected spiking neural network allows reducing the number of connections by up to 60% compared to a fully connected architecture. At the same time, the accuracy of solving the classification problem does not deteriorate and is 0.92...0.95 according to the F1 metric. This matches the accuracy of standard support vector machine, k-nearest neighbor, and random forest classifiers. The source code for this article is publicly available: https://github.com/sag111/Sparse-WTA-SNN.