SPARTA: Advancing Sparse Attention in Spiking Neural Networks via Spike-Timing-Based Prioritization

TL;DR

SPARTA introduces a spike-timing-based sparse attention framework for spiking neural networks, significantly reducing computational complexity while maintaining high accuracy by prioritizing salient tokens through temporal cues.

Contribution

It presents a novel method leveraging spike timing for efficient sparse attention in SNNs, improving computational efficiency and accuracy over prior rate-based approaches.

Findings

Achieves 65.4% sparsity with competitive gating.

Reduces attention complexity from O(N^2) to O(K^2).

State-of-the-art 98.78% on DVS-Gesture.

Abstract

Current Spiking Neural Networks (SNNs) underutilize the temporal dynamics inherent in spike-based processing, relying primarily on rate coding while overlooking precise timing information that provides rich computational cues. We propose SPARTA (Spiking Priority Attention with Resource-Adaptive Temporal Allocation), a framework that leverages heterogeneous neuron dynamics and spike-timing information to enable efficient sparse attention. SPARTA prioritizes tokens based on temporal cues, including firing patterns, spike timing, and inter-spike intervals, achieving 65.4% sparsity through competitive gating. By selecting only the most salient tokens, SPARTA reduces attention complexity from O(N^2) to O(K^2) with k << n, while maintaining high accuracy. Our method achieves state-of-the-art performance on DVS-Gesture (98.78%) and competitive results on CIFAR10-DVS (83.06%) and CIFAR-10 (95.3%), demonstrating that exploiting spike timing dynamics improves both computational efficiency and accuracy.