New method trains energy-efficient spiking neural networks faster

Researchers have developed EGGROLL, a novel gradient-free method for training Spiking Neural Networks (SNNs) that significantly reduces computational cost. This approach uses low-rank factorization of Evolution Strategies perturbations, decreasing memory requirements from O(mn) to O(r(m+n)). Applied to the N-MNIST dataset, EGGROLL achieved 79.21% test accuracy while speeding up training by 2.23x compared to full-rank Evolution Strategies, making it suitable for neuromorphic hardware without surrogate gradients. AI

IMPACT Enables more efficient training of SNNs for neuromorphic hardware, potentially leading to lower power consumption in AI applications.

RANK_REASON The cluster contains an academic paper detailing a new method for training SNNs. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

• EGGROLL
• Evolution Strategies
• N-MNIST
• Spiking Neural Networks

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →