用主动预测过滤提升脉冲变压器的能效与精度
SAFformer:Improving Spiking Transformer via Active Predictive Filtering
- 基于大脑预测编码机制，主动抑制可预测信号
- ImageNet-1K上达80.44%准确率，仅26.58M参数
- 适合低功耗视觉推理场景，尤其注重能效比
脉冲神经网络（SNNs）在生物合理性与能效方面具有显著优势，是构建低功耗Transformer的有前景候选。然而，现有脉冲Transformer大多遵循被动反应范式，难以聚焦任务相关信息，处理冗余视觉数据时计算开销大。为此，我们提出SAFformer，一种基于主动预测过滤范式的新型脉冲Transformer架构。受大脑预测编码机制启发，SAFformer主动抑制可预测信号，专注于显著视觉特征。大量实验表明，SAFformer在CIFAR-10/100和CIFAR10-DVS上达到新最优性能。尤为突出的是，在ImageNet-1K上，其Top-1准确率达80.44%，参数量仅26.58M，能耗为5.88 mJ，展现出优异的精度与效率平衡。
原文摘要 · Abstract (English)
Spiking Neural Networks (SNNs) offer notable advantages in biological plausibility and energy efficiency, making them promising candidates for building low-power Transformers. However, existing Spiking Transformers largely adhere to a passive reactive paradigm, which struggles to focus on task-relevant information and incurs substantial computational overhead when processing redundant visual data. To overcome this fundamental yet underexplored limitation, we propose SAFformer, a novel Spiking Transformer architecture based on an active predictive filtering paradigm. Inspired by the brain's predictive coding mechanism, SAFformer actively suppresses predictable signals and focuses on salient visual features. Extensive experiments show that SAFformer establishes new state-of-the-art performance on CIFAR-10/100 and CIFAR10-DVS. Remarkably, on ImageNet-1K, it achieves 80.44% Top-1 accuracy with only 26.58M parameters and an energy consumption of 5.88 mJ, demonstrating an exceptional balance between accuracy and efficiency.
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。