Персона: Рыбка, Роман Борисович
Загружается...
Email Address
Birth Date
Научные группы
Организационные подразделения
Организационная единица
Институт лазерных и плазменных технологий
Стратегическая цель Института ЛаПлаз – стать ведущей научной школой и ядром развития инноваций по лазерным, плазменным, радиационным и ускорительным технологиям, с уникальными образовательными программами, востребованными на российском и мировом рынке образовательных услуг.
Статус
Фамилия
Рыбка
Имя
Роман Борисович
Имя
4 results
Результаты поиска
Теперь показываю 1 - 4 из 4
- ПубликацияТолько метаданныеNeural-network method for determining text author's sentiment to an aspect specified by the named entity(2020) Naumov, A.; Rybka, R.; Sboev, A.; Selivanov, A.; Gryaznov, A.; Рыбка, Роман Борисович; Сбоев, Александр Георгиевич© 2020 Copyright for this paper by its authors.This study presents the approach to aspect-based sentiment analysis where a named entity of a certain category is considered as an aspect. Such task formulation is a novelty and opens up the opportunity to determine writers' attitudes to organizations and people considered in texts. This task required a dataset of Russian-language sentences where sentiment with respect to certain named entities would be labeled, which we collected using a crowdsourcing platform. Sentiment determination is based on a deep neural network with attention mechanism and ELMo language model for word vector representation. The proposed model is validated on available data on a similar task. The resulting performance (by the f1-micro metric) on the collected dataset is 0.72, which is the new state of the art for the Russian language.
- ПубликацияТолько метаданныеEnsembling SNNs with STDP Learning on Base of Rate Stabilization for Image Classification(2021) Serenko, A.; Sboev, A.; Rybka, R.; Vlasov, D.; Сбоев, Александр Георгиевич; Рыбка, Роман Борисович© 2021, The Author(s), under exclusive license to Springer Nature Switzerland AG.In spite of a number of existing spiking neural network models for image classification, it still remains relevant from both methodological and practical points of view to develop a model as simple as possible, while at the same time applicable to classification tasks with various types of data, be it real vectors or images. Our previous work proposed a simple spiking network with Spike-Timing-Dependent-Plasticity (STDP) learning for solving real-vector classification tasks. In this paper, that method is extended to image recognition tasks and enhanced by aggregating neurons into ensembles. The network comprises one layer of neurons with STDP-plastic inputs receiving pixels of input images encoded with spiking rates. This work considers two approaches for aggregating neurons’ output activities within an ensemble: by averaging their output spiking rates (i.e. averaging outputs before decoding spiking rates into class labels) and by voting with decoded class labels. Ensembles aggregated by output frequencies are shown to achieve a significant accuracy increase up to 95% (by F1-score) for the Optdigits handwritten digit dataset, and is comparable with conventional machine learning approaches.
- ПубликацияТолько метаданныеImage and Audio Data Classification Using Bagging Ensembles of Spiking Neural Networks with Memristive Plasticity(2024) Rybka, R. B.; Davydov, Y.; Sboev, A.; Vlasov, D.; Рыбка, Роман Борисович; Сбоев, Александр Георгиевич
- ПубликацияТолько метаданныеDeep 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.