Optimization and application of dendritic neuron model

Abstract

As the well-known McCulloch–Pitts neuron model has long been criticized for being oversimplified, different algebra to formulate a single neuron model has received increasing interest. The dendritic neuron model (DNM) which considers the nonlinear information processing capacity of both synapses and dendrites has shown its effectiveness on classification problems. However, an effective learning method for DNM is still highly desired and challenging because the traditional error back-propagation (BP) algorithm usually suffers from issues originating from the proliferation of saddle points and local minima trapping.

In addition to BP, the mainstream DNM optimization methods include meta-heuristic algorithms (MHAs). However, over the decades, MHAs have developed a large number of different algorithms. How to screen suitable MHAs for optimizing DNM has become a hot and challenging area of research. In this study, we classified MHAs into different clusters with different population interaction networks (PIN). The performance of DNMs optimized by different clusters of MHAs is tested in the prediction and classification tasks.

Date
Nov 22, 2023 2:41 AM — 2:44 AM
Event
Group Research
Location
WeAI Research Group & University of Toyama

Researchers:

Jiayi Li 李加一

Yuxin Zhang 張雨馨

Yifei Yang 楊逸飛

Xiaosi Li 李小司

Haichuan Yang 楊海川

Published Journals:

  1. Wang, Z., Gao, S., Wang, J., Yang, H., & Todo, Y. (2020). A dendritic neuron model with adaptive synapses trained by differential evolution algorithm. Computational intelligence and neuroscience, 2020.
  2. Yang, H., Yu, Y., Cheng, J., Lei, Z., Cai, Z., Zhang, Z., & Gao, S. (2022). An intelligent metaphor-free spatial information sampling algorithm for balancing exploitation and exploration. Knowledge-Based Systems, 250, 109081.
  3. Xu, Z., Wang, Z., Li, J., Jin, T., Meng, X., & Gao, S. (2021). Dendritic neuron model trained by information feedback-enhanced differential evolution algorithm for classification. Knowledge-Based Systems, 233, 107536.
  4. Yang Y, Li X, Li H, Zhang C, Todo Y, Yang H. Yet Another Effective Dendritic Neuron Model Based on the Activity of Excitation and Inhibition. Mathematics. 2023; 11(7):1701. Code
  5. Zhang Y, Yang Y, Li X, Yuan Z, Todo Y, Yang H. A Dendritic Neuron Model Optimized by Meta-Heuristics with a Power-Law-Distributed Population Interaction Network for Financial Time-Series Forecasting. Mathematics. 2023; 11(5):1251. Code
  6. Yang, H., Zhang, Y., Zhang, C., Xia, W., Yang, Y., & Zhang, Z. (2023). A Hyperparameter Self-Evolving SHADE-Based Dendritic Neuron Model for Classification. Axioms, 12(11), 1051. Code
Haichuan Yang 楊海川
Haichuan Yang 楊海川
assistant professor

My research interests include meta-heuristic algorithms, artificial neuron model and complex systems.