Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

CogniSNN: An Exploration to Random Graph Architecture based Spiking Neural Networks with Enhanced Depth-Scalability and Path-Plasticity

Yongsheng Huang, Peibo Duan, Zhipeng Liu, Kai Sun, Changsheng Zhang, Bin Zhang, Mingkun Xu

TL;DR

CogniSNN introduces a cognition-aware SNN built on a random graph architecture to achieve depth-scalability and path-plasticity. It combines ResNode-based random graphs with an OR-based skip to preserve spike computation, a tailored pooling mechanism to fix feature misalignment, and a critical-path LwF algorithm guided by betweenness centrality to facilitate continual learning. Experimental results on neuromorphic datasets show competitive accuracy with far fewer parameters and strong continual-learning performance, along with favorable energy characteristics. The work demonstrates the feasibility of RGA-based SNNs for brain-inspired depth and path adaptation and points to future graph-structured, adaptive spiking systems.

Abstract

Currently, most spiking neural networks (SNNs) still mimic the chain-like hierarchical architecture in traditional artificial neural networks (ANNs). This method significantly differs from random connections between neurons found in biological brains, limiting the ability to model the evolving mechanisms of neural pathways in biological neural systems, particularly in terms of dynamic depth-scalability and adaptive path-plasticity. This paper develops a new modeling paradigm for SNNs with random graph architecture (RGA), termed Cognition-aware SNN (CogniSNN). Furthermore, we model the depth-scalability and path-plasticity in CogniSNN by introducing a modified spiking residual neural node (ResNode) to counteract network degradation in deeper graph pathways, as well as a critical path-based algorithm that enables CogniSNN to perform path reusability on new tasks leveraging the features of the data and the RGA learned in old tasks. Experiments show that the performance of CogniSNN with redesigned ResNode is comparable, even superior, to current state-of-the-art SNNs on neuromorphic datasets. The critical path-based approach effectively achieves path reuse capability while maintaining expected performance in learning new tasks that are similar to or distinct from the old ones. This study showcases the potential of RGA-based SNNs and paves a new path for modeling the fusion of computational neuroscience and deep intelligent agents. The code is available at github.com/Yongsheng124/CogniSNN.

CogniSNN: An Exploration to Random Graph Architecture based Spiking Neural Networks with Enhanced Depth-Scalability and Path-Plasticity

TL;DR

CogniSNN introduces a cognition-aware SNN built on a random graph architecture to achieve depth-scalability and path-plasticity. It combines ResNode-based random graphs with an OR-based skip to preserve spike computation, a tailored pooling mechanism to fix feature misalignment, and a critical-path LwF algorithm guided by betweenness centrality to facilitate continual learning. Experimental results on neuromorphic datasets show competitive accuracy with far fewer parameters and strong continual-learning performance, along with favorable energy characteristics. The work demonstrates the feasibility of RGA-based SNNs for brain-inspired depth and path adaptation and points to future graph-structured, adaptive spiking systems.

Abstract

Currently, most spiking neural networks (SNNs) still mimic the chain-like hierarchical architecture in traditional artificial neural networks (ANNs). This method significantly differs from random connections between neurons found in biological brains, limiting the ability to model the evolving mechanisms of neural pathways in biological neural systems, particularly in terms of dynamic depth-scalability and adaptive path-plasticity. This paper develops a new modeling paradigm for SNNs with random graph architecture (RGA), termed Cognition-aware SNN (CogniSNN). Furthermore, we model the depth-scalability and path-plasticity in CogniSNN by introducing a modified spiking residual neural node (ResNode) to counteract network degradation in deeper graph pathways, as well as a critical path-based algorithm that enables CogniSNN to perform path reusability on new tasks leveraging the features of the data and the RGA learned in old tasks. Experiments show that the performance of CogniSNN with redesigned ResNode is comparable, even superior, to current state-of-the-art SNNs on neuromorphic datasets. The critical path-based approach effectively achieves path reuse capability while maintaining expected performance in learning new tasks that are similar to or distinct from the old ones. This study showcases the potential of RGA-based SNNs and paves a new path for modeling the fusion of computational neuroscience and deep intelligent agents. The code is available at github.com/Yongsheng124/CogniSNN.
Paper Structure (20 sections, 2 theorems, 14 equations, 3 figures, 6 tables, 1 algorithm)

This paper contains 20 sections, 2 theorems, 14 equations, 3 figures, 6 tables, 1 algorithm.

Table of Contents

  1. Related Work
  2. Residual Structures in SNNs
  3. Continual Learning in SNNs
  4. Preliminaries
  5. Spiking Neuron
  6. Learning without Forgetting
  7. Modeling and Methodology
  8. Modeling of CogniSNN
  9. ResNode
  10. Tailored Pooling Mechanism
  11. Critical Path-Based LwF
  12. Experiments
  13. Experimental Setting
  14. Preliminary Analysis
  15. Insightful Performance Analysis
  16. ...and 5 more sections

Key Result

Theorem 1

The operation of an OR gate performs an identity mapping function.

Figures (3)

  • Figure 1: Comprehensive diagram of CogniSNN. During neuromorphic object recognition, data flows from the $ConvBNSN$ triple to the RGA-based SNN and then enters the classifier. The internal structure of ResNode and the illustration of dimension misalignment are shown at the bottom right and right side of the figure, respectively. During continual learning, different critical paths are selected for learning new knowledge in different scenarios.
  • Figure 2: Comparison of results on three neuromorphic datasets between CogniSNN and the chain-like structured SNN.
  • Figure 3: Comparison of results of OR and other operations on DVS-Gesture with ER-RGA-7 based CogniSNN.

Theorems & Definitions (5)

  • Theorem 1
  • proof
  • Corollary 2
  • Definition 1
  • Definition 2