RADIAL BASIS NEURAL NETWORK WITH MULTIPLE CONNECTED WEIGHTS
Дата публикации
25.04.2026
Журнал
Приоритетные области применения искусственного интеллекта в педагогическом образовании
Выпуск
Приоритетные области применения искусственного интеллекта в педагогическом образовании
Страницы
344-350
Авторы
Аннотация
In this work, we propose a new type of radial basis neural network model where the connection between two units is not a single value but a set of values, which means multi-connected weights exist. In our model, each pair of units is connected by more than one link. These links mimic different neurotransmitters, and their number reflects the number of neurotransmitter types considered. Experimental tests on benchmark datasets from the Machine Learning Repository show that using radial basis with multiple weight connections improves performance over traditional neural networks. This method gives a new way to design and build artificial neural networks.
Ключевые слова
classification
radial basis function
radial basis neural network
neurotransmitter
multiple connections
weight
hidden layer
Список литературы
1. Y. Yin, Z. Han, M. Jian, G. Wang, L. Chen, R. Wang. AMSUnet: A neural network using atrous multi-scale convolution for medical image segmentation. Computers in Biology and Medicine. Volume 162, August 2023, 107120, doi:10.1016/j.compbiomed.2023.107120.
2. M. Farrell, S. Recanatesi, E. Brown. From lazy to rich to exclusive task representations in neural networks and neural codes. Current Opinion in Neurobiology. 2023, Vol. 83, 102780, doi:10.1016/j.conb.2023.102780.
3. A. Apicella, F. Isgrò, A. Pollastro, R. Prevete. Adaptive filters in Graph Convolutional Neural Networks. Pattern Recognition. 2023, Vol. 144, 109867, doi:10.1016/j.patcog.2023.109867.