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.
Abstract (translated)
当前的脉冲神经网络（SNNs）未能充分利用基于脉冲处理中固有的时间动态特性，主要依赖于频率编码，而忽略了提供丰富计算线索的确切时间信息。我们提出了SPARTA（具有资源自适应时序分配的脉冲优先注意机制），这是一个框架，它利用异构神经元动力学和脉冲时间信息来实现高效的稀疏注意力。SPARTA 通过包括放电模式、脉冲时间和脉冲间间隔的时间线索对令牌进行优先级排序，并通过竞争性门控实现了65.4%的稀疏度。通过仅选择最重要的令牌，SPARTA 将注意力复杂度从 O(N^2) 减少到 O(K^2)，其中 k << n，同时保持高精度。 我们的方法在 DVS-Gesture 数据集上达到了 98.78% 的最佳性能，并且在 CIFAR10-DVS（83.06%）和 CIFAR-10（95.3%）数据集上取得了有竞争力的结果。这表明利用脉冲时间动力学可以提高计算效率并改善准确性。
URL
https://arxiv.org/abs/2508.01646
