Designing neural networks based on dendritic neuron model
Abstract
The dendritic neuron model (DNM), which considers the nonlinear information processing capability of synapses and dendrites, has more robust performance and interpretability than the McCulloch-Pitts neuron model. However, most research on DNM has been limited to single neurons. How to use DNMs to form neural networks and solve real-world problems is a complex problem to be solved.
At present, we use ResNet, which is very powerful in deep learning, to combine with DNM to design DResNet with stronger performance. DResNet has shown powerful ability in COVID-19 detection problems. It can accurately detect patients with COVID-19 with over 98% accuracy. In the future, we will realize the use of DNM to compose more complex neural networks and apply them to various hard-to-solve practical problems.
Researchers:
Published Journals:
- Li J, Liu Z, Wang R L, et al. Dendritic Deep Residual Learning for COVID‐19 Prediction[J]. IEEJ Transactions on Electrical and Electronic Engineering, 2023, 18(2): 297-299.
Haichuan Yang 楊海川
assistant professor
My research interests include meta-heuristic algorithms, artificial neuron model and complex systems.