通过尖峰时序优先级实现稀疏化注意力，提升脉冲神经网络效率与精度。
SPARTA: Advancing Sparse Attention in Spiking Neural Networks via Spike-Timing-Based Prioritization
- 基于尖峰时序与发放模式动态分配注意力优先级
- 实现65.4%稀疏度，计算复杂度从O(N²)降至O(K²)
- 适用于需高效低功耗的实时视觉任务场景
当前脉冲神经网络（SNNs）未能充分利用尖峰信号中的时间动态特性，主要依赖率编码而忽视了提供丰富计算线索的精确时序信息。本文提出SPARTA（基于尖峰时序的脉冲优先注意力与资源自适应时间分配），利用异构神经元动态和尖峰时序信息实现高效的稀疏注意力。该方法依据放电模式、尖峰时序及峰间间隔等时间线索对输入标记进行优先级排序，通过竞争性门控实现65.4%的稀疏度。通过仅选择最显著的标记，将注意力复杂度从O(N²)降低至O(K²)，其中K远小于N，同时保持高精度。在DVS-Gesture上达到98.78%准确率，在CIFAR10-DVS上取得83.06%的竞争力结果，在CIFAR-10上达95.3%，证明挖掘尖峰时序动态能同时提升计算效率与分类性能。
原文摘要 · Abstract (English)
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.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。