Персона: Сбоев, Александр Георгиевич
Загружается...
Email Address
Birth Date
Научные группы
Организационные подразделения
Организационная единица
Институт ядерной физики и технологий
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.
Статус
Фамилия
Сбоев
Имя
Александр Георгиевич
Имя
4 results
Результаты поиска
Теперь показываю 1 - 4 из 4
- ПубликацияОткрытый доступComparison of Bagging and Sparcity Methods for Connectivity Reduction in Spiking Neural Networks with Memristive Plasticity(2024) Rybka, R.; Davydov, Y.; Vlasov, D.; Sboev, A.; Рыбка, Роман Борисович; Сбоев, Александр ГеоргиевичDeveloping a spiking neural network architecture that could prospectively be trained on energy-efficient neuromorphic hardware to solve various data analysis tasks requires satisfying the limitations of prospective analog or digital hardware, i.e., local learning and limited numbers of connections, respectively. In this work, we compare two methods of connectivity reduction that are applicable to spiking networks with local plasticity; instead of a large fully-connected network (which is used as the baseline for comparison), we employ either an ensemble of independent small networks or a network with probabilistic sparse connectivity. We evaluate both of these methods with a three-layer spiking neural network, which are applied to handwritten and spoken digit classification tasks using two memristive plasticity models and the classical spike time-dependent plasticity (STDP) rule. Both methods achieve an F1-score of 0.93ў??0.95 on the handwritten digits recognition task and 0.85ў??0.93 on the spoken digits recognition task. Applying a combination of both methods made it possible to obtain highly accurate models while reducing the number of connections by more than three times compared to the basic model.
- ПубликацияТолько метаданныеMemristor-based spiking neural network with online reinforcement learning(2023) Vlasov, D.; Minnekhanov, A.; Rybka, R.; Davydov, Y.; Sboev, A. G.; Сбоев, Александр ГеоргиевичNeural 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
- ПубликацияОткрытый доступExtraction of Significant Features by Fixed-Weight Layer of Processing Elements for the Development of an Efficient Spiking Neural Network Classifier(2023) Sboev, A.; Rybka, R.; Kunitsyn, D.; Serenko, A.; Сбоев, Александр ГеоргиевичIn this paper, we demonstrate that fixed-weight layers generated from random distribution or logistic functions can effectively extract significant features from input data, resulting in high accuracy on a variety of tasks, including Fisher’s Iris, Wisconsin Breast Cancer, and MNIST datasets. We have observed that logistic functions yield high accuracy with less dispersion in results. We have also assessed the precision of our approach under conditions of minimizing the number of spikes generated in the network. It is practically useful for reducing energy consumption in spiking neural networks. Our findings reveal that the proposed method demonstrates the highest accuracy on Fisher’s iris and MNIST datasets with decoding using logistic regression. Furthermore, they surpass the accuracy of the conventional (non-spiking) approach using only logistic regression in the case of Wisconsin Breast Cancer. We have also investigated the impact of non-stochastic spike generation on accuracy.
- ПубликацияТолько метаданныеSpiking 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 ГеоргиевичPurpose. 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.