SAFformer:Improving Spiking Transformer via Active Predictive Filtering
Researchers have introduced SAFformer, a novel Spiking Transformer architecture designed to improve energy efficiency and performance in visual data processing. By adopting an active predictive filtering paradigm inspired by the brain's predictive coding, SAFformer actively suppresses predictable signals and concentrates on salient visual features. This approach has led to new state-of-the-art results on CIFAR-10/100 and CIFAR10-DVS datasets, and achieved notable accuracy with significantly reduced parameters and energy consumption on ImageNet-1K. AI
IMPACT This new architecture demonstrates a significant improvement in energy efficiency for visual processing tasks, potentially enabling more powerful AI on low-power devices.