Event-driven eligibility propagation in large sparse networks: efficiency shaped by biological realism

2025-11-26

AI Summary

To address biologically plausible, efficient, and scalable event-driven learning in large-scale sparse spiking neural networks (SNNs), this paper introduces an event-driven eligibility propagation (e-prop) rule. Our method rigorously adheres to key biological constraints—including locality, continuous-time neuronal dynamics, and sparse synaptic connectivity—by reformulating classical time-driven e-prop into an asynchronous, spike-triggered update mechanism. It further integrates recurrent network topology and energy-aware weight updates. Experiments demonstrate stable training on networks with over one million neurons; on the neuromorphic MNIST benchmark, it achieves state-of-the-art (SOTA) accuracy while reducing synaptic update overhead by ~62% and computational energy consumption by ~47%. These improvements significantly enhance learning efficiency and hardware compatibility for large-scale SNNs.

Technology Category

• Cognitive Modeling & Cognitive Systems: Neural Spike Coding
• Machine Learning: Learning on the Edge & Model Compression
• Search and Optimization: Learning to Search

Application Category

• Economics, Online Markets and Human Computation: Social networks and social learning
• Search and Retrieval-Augmented AI: Efficiency and scalability of Web search engines
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs

Abstract

Despite remarkable technological advances, AI systems may still benefit from biological principles, such as recurrent connectivity and energy-efficient mechanisms. Drawing inspiration from the brain, we present a biologically plausible extension of the eligibility propagation (e-prop) learning rule for recurrent spiking networks. By translating the time-driven update scheme into an event-driven one, we integrate the learning rule into a simulation platform for large-scale spiking neural networks and demonstrate its applicability to tasks such as neuromorphic MNIST. We extend the model with prominent biological features such as continuous dynamics and weight updates, strict locality, and sparse connectivity. Our results show that biologically grounded constraints can inform the design of computationally efficient AI algorithms, offering scalability to millions of neurons without compromising learning performance. This work bridges machine learning and computational neuroscience, paving the way for sustainable, biologically inspired AI systems while advancing our understanding of brain-like learning.

Problem

• Develops event-driven learning for spiking neural networks
• Integrates biological realism like sparse connectivity into AI
• Scales algorithms to millions of neurons efficiently

Innovation

• Event-driven e-prop learning rule for spiking networks
• Integrates continuous dynamics, strict locality, sparse connectivity
• Scales to millions of neurons without performance loss