Article
Computer Science, Artificial Intelligence
Haixu Ding, Wenjing Li, Junfei Qiao
Summary: The paper proposes a self-organizing recurrent fuzzy neural network based on multivariate time series analysis, which utilizes a recurrent mechanism and a self-organization mechanism to optimize network structure. The theoretical analysis of its convergence and practical validation demonstrate its effectiveness in modeling nonlinear systems.
NEURAL COMPUTING & APPLICATIONS
(2021)
Article
Biology
Yue Kris Wu, Friedemann Zenke
Summary: Neural circuits can achieve rapid information processing through nonlinear transient amplification, which involves two phases - selective amplification of inputs exceeding a critical threshold by positive feedback excitation, and stabilization of runaway dynamics into an inhibitory state by short-term plasticity. NTA offers a parsimonious explanation for how excitatory-inhibitory co-tuning and short-term plasticity collaborate in recurrent networks to achieve transient amplification.
Article
Computer Science, Artificial Intelligence
Yin Su, Cuili Yang, Junfei Qiao
Summary: The self-organizing PRWNN (SPRWNN) is a neural network model that automatically adjusts the network structure by using the spiking strength of nodes and the module growth mechanism. Experimental results show that SPRWNN improves the prediction accuracy by about 40% compared to the PRWNN with fixed structure.
EXPERT SYSTEMS WITH APPLICATIONS
(2023)
Article
Mathematics, Interdisciplinary Applications
Wei Zhou, Bowei Fu, Guangyi Wang
Summary: This paper focuses on the hybrid effects of memristor characteristics, coupling coefficient, and time-delay on the recurrent neural network. A novel time-delay recurrent neural network based on a passive hyperbolic tangent memristor is constructed, and its dynamic behaviors are analyzed. The theoretical results are verified through circuit simulation.
INTERNATIONAL JOURNAL OF BIFURCATION AND CHAOS
(2022)
Article
Metallurgy & Metallurgical Engineering
Samira Johari, Mahdi Yaghoobi, Hamid R. Kobravi
Summary: This paper introduces a nonlinear model predictive controller (NMPC) based on hyper chaotic diagonal recurrent neural network (HCDRNN) for modeling and predicting the behavior of under-controlled systems. The proposed method demonstrates superior performance in trajectory tracking and disturbance rejection, with advantages including parameter convergence, negligible prediction error, guaranteed stability, and high tracking performance.
JOURNAL OF CENTRAL SOUTH UNIVERSITY
(2022)
Article
Physics, Multidisciplinary
Maria Masoliver, Jorn Davidsen, Wilten Nicola
Summary: This study demonstrates the generation of embedded chimera states in recurrent neural networks through machine learning. These states can be produced by networks with connectivity matrices slightly perturbed from random networks, and exhibit robustness.
COMMUNICATIONS PHYSICS
(2022)
Article
Computer Science, Artificial Intelligence
Gongming Wang, Junfei Qiao
Summary: This article presents an efficient self-organizing fuzzy neural network (SOFNN) called IDPT-SOFNN, which is capable of extracting effective features and dynamically adjusting its structure for better learning speed, accuracy, and generalization capability. It has shown superior performance compared to existing methods in handling practical complex data.
IEEE TRANSACTIONS ON FUZZY SYSTEMS
(2022)
Article
Mathematics, Interdisciplinary Applications
Eunice Leung, King F. Ma, Nan Xie
Summary: In this paper, the bubble sound of different types of sparkling drinks is collected and a data driven model for bubble dynamics is constructed using experimental data. It is verified that the bubble dynamics of sparkling drinks are nonlinear and chaotic. The proposed PI-LSTM neural network, guided by the derived physical principle on bubble pressure, shows improved accuracy in modeling the bubble sound data compared to standard predictors.
CHAOS SOLITONS & FRACTALS
(2023)
Article
Engineering, Marine
Ning Wang, Huihui Wu, Yuhang Zhang, Jialin Song, Yejin Lin, Lizhu Hao
Summary: This paper introduces a self-organizing data-driven network with hierarchical pruning model using fuzzy neural network for fast-dynamics prediction in ship maneuvering. Through incremental training and hierarchical pruning mechanism, the model is able to accurately predict the velocity dynamics of the ship and achieve dynamic abstraction.
Article
Physics, Multidisciplinary
Alexander van Meegen, Tobias Kuehn, Moritz Helias
Summary: This study unifies the field-theoretical approach to neuronal networks with large deviations theory, deriving a rate function resembling Kullback-Leibler divergence through field theory to enable data-driven inference of model parameters and calculation of fluctuations beyond mean-field theory. Additionally, the study reveals a regime with fluctuation-induced transitions between mean-field solutions.
PHYSICAL REVIEW LETTERS
(2021)
Article
Computer Science, Artificial Intelligence
Juncheng Zhang, Fei Chao, Hualin Zeng, Chih-Min Lin, Longzhi Yang
Summary: This paper introduces a novel brain emotional neural network that integrates a wavelet neural network into a conventional brain emotional learning network and employs a recurrent structure to address the challenges of nonlinearity and uncertainty in control systems. The proposed network outperformed other popular neural-network-based control systems in experiments on uncertain nonlinear systems.
Article
Engineering, Multidisciplinary
Augusto Cheffer, Marcelo A. Savi, Tiago Leite Pereira, Aline Souza de Paula
Summary: This study investigates cardiac rhythms through a mathematical model analyzing the electrical activity of the heart and conducting numerical simulations to reproduce synthetic ECGs with various responses. The results show that nonlinear tools can assist in understanding physiology and characterizing pathologies.
APPLIED MATHEMATICAL MODELLING
(2021)
Article
Mathematics
Ali Najem Alkawaz, Jeevan Kanesan, Irfan Anjum Badruddin, Sarfaraz Kamangar, Mohamed Hussien, Maughal Ahmed Ali Baig, N. Ameer Ahammad
Summary: This study presents two models of self-organizing map (SOM) formulated as an optimal control problem. The first model focuses on the weight updating equation of the best matching units in each iteration, while the second model considers the weight updating equation of all nodes in the SOM. The SOMOC2 model performs better by considering all nodes in the Hamiltonian equation and produces a greater improvement in terms of mean quantization error.
Article
Engineering, Electrical & Electronic
Fatemeh Charoosaei, Amin Faraji, Sayed Alireza Sadrossadat, Ali Mirvakili, Weicong Na, Feng Feng, Qi-Jun Zhang
Summary: In the field of computer-aided design (CAD), the use of recurrent neural networks (RNN) has proven to be highly effective in generating fast and high-performance models. One key challenge in this area is predicting time sequences, which requires identifying the dependencies between sequences. Conventional RNNs face limitations in terms of accuracy and the number of parameters. To address this, we propose a new macromodeling method called Clockwork-RNN (CWRNN) and its hybrid version, which simplifies the architecture and reduces model complexity while still accurately capturing complex dependencies. The CWRNN also offers lower computational cost and greater flexibility in architectural configuration.
IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS I-REGULAR PAPERS
(2023)
Article
Optics
Mehdi Mabed, Fanchao Meng, Lauri Salmela, Christophe Finot, Goery Genty, John M. Dudley
Summary: This study demonstrates that neural networks can accurately predict the time-domain properties of optical fiber instabilities by analyzing spectral intensity profiles. It extends the previous research on machine learning prediction for single-pass fiber propagation instabilities to the more complex case of noise-like pulse dynamics in a dissipative soliton laser.