LE QUY DON
Technical University
VietnameseClear Cookie - decide language by browser settings

Solutions to Improve the Performance of the Algorithm with the Adaptive Decay Time for the Spiking Neural Nets

Khoa, T.D. and Tuan, N.V. and Dung, P.T. and Trang, N.T.T. and Thanh, N.D. (2024) Solutions to Improve the Performance of the Algorithm with the Adaptive Decay Time for the Spiking Neural Nets. In: International Conference on Intelligent Systems and Networks, ICISN 2024, 22 March 2024 Through 23 March 2024, Hanoi.

Full text not available from this repository. (Upload)

Abstract

This paper introduces an adaptation of the learning rate and momentum for the backpropagation algorithm during the training spiking neural network with the decay time of the spike response model. The efficiency of the algorithm with the proposed solutions is compared with the original algorithm through the XOR classification problem and the aerodynamic coefficients identification of an aircraft from the data sets recorded from flights. The results show that the algorithm with the proposed solutions has a higher successful classification rate on the XOR problem as well as higher accuracy in the aerodynamic coefficients identification problem compared to the original algorithm, whereas the number of epochs is much smaller. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024.

Item Type: Conference or Workshop Item (Paper)
Divisions: Faculties > Faculty of Control Engineering
Identification Number: 10.1007/978-981-97-5504-2₁₂
Uncontrolled Keywords: Backpropagation; Training aircraft, Aerodynamic coefficients; Learning rates; Neural-networks; Original algorithms; Performance; Spike response models; Spiking neural nets; Spiking neural network; The decay time, Neural network models
Additional Information: Conference of International Conference on Intelligent Systems and Networks, ICISN 2024 ; Conference Date: 22 March 2024 Through 23 March 2024; Conference Code:318189
URI: http://eprints.lqdtu.edu.vn/id/eprint/11375

Actions (login required)

View Item
View Item