TY - JOUR
T1 - Fully Complex-Valued Dendritic Neuron Model
AU - Gao, Shangce
AU - Zhou, Meng Chu
AU - Wang, Ziqian
AU - Sugiyama, Daiki
AU - Cheng, Jiujun
AU - Wang, Jiahai
AU - Todo, Yuki
N1 - Funding Information:
This work was supported in part by the National Key Research and Development Program of China under Grant 2018AAA0101203, in part by the National Natural Science Foundation of China under Grant 61872271 and Grant 62072483
Publisher Copyright:
© 2012 IEEE.
PY - 2023/4/1
Y1 - 2023/4/1
N2 - A single dendritic neuron model (DNM) that owns the nonlinear information processing ability of dendrites has been widely used for classification and prediction. Complex-valued neural networks that consist of a number of multiple/deep-layer McCulloch-Pitts neurons have achieved great successes so far since neural computing was utilized for signal processing. Yet no complex value representations appear in single neuron architectures. In this article, we first extend DNM from a real-value domain to a complex-valued one. Performance of complex-valued DNM (CDNM) is evaluated through a complex XOR problem, a non-minimum phase equalization problem, and a real-world wind prediction task. Also, a comparative analysis on a set of elementary transcendental functions as an activation function is implemented and preparatory experiments are carried out for determining hyperparameters. The experimental results indicate that the proposed CDNM significantly outperforms real-valued DNM, complex-valued multi-layer perceptron, and other complex-valued neuron models.
AB - A single dendritic neuron model (DNM) that owns the nonlinear information processing ability of dendrites has been widely used for classification and prediction. Complex-valued neural networks that consist of a number of multiple/deep-layer McCulloch-Pitts neurons have achieved great successes so far since neural computing was utilized for signal processing. Yet no complex value representations appear in single neuron architectures. In this article, we first extend DNM from a real-value domain to a complex-valued one. Performance of complex-valued DNM (CDNM) is evaluated through a complex XOR problem, a non-minimum phase equalization problem, and a real-world wind prediction task. Also, a comparative analysis on a set of elementary transcendental functions as an activation function is implemented and preparatory experiments are carried out for determining hyperparameters. The experimental results indicate that the proposed CDNM significantly outperforms real-valued DNM, complex-valued multi-layer perceptron, and other complex-valued neuron models.
KW - Activation functions
KW - McCulloch-Pitts neuron
KW - complex back-propagation (BP)
KW - complex domain
KW - complex-valued neural networks
KW - dendritic neuron model (DNM)
KW - elementary transcendental functions
UR - http://www.scopus.com/inward/record.url?scp=85114727453&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85114727453&partnerID=8YFLogxK
U2 - 10.1109/TNNLS.2021.3105901
DO - 10.1109/TNNLS.2021.3105901
M3 - Article
C2 - 34487498
AN - SCOPUS:85114727453
SN - 2162-237X
VL - 34
SP - 2105
EP - 2118
JO - IEEE Transactions on Neural Networks and Learning Systems
JF - IEEE Transactions on Neural Networks and Learning Systems
IS - 4
ER -