A signed quantized activation plus a multi-spike signed neuron makes a one-timestep spiking network reproduce the quantized ANN exactly, giving ANN-identical accuracy at T=1 on CIFAR-10/100 and state-of-the-art one-step accuracy on ImageNet and event datasets.
Spikedattention: Training-free and fully spike-driven transformer-to-SNN conversion with winner-oriented spike shift for softmax operation,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
baseline 1
citation-polarity summary
fields
cs.NE 1years
2025 1verdicts
CONDITIONAL 1roles
baseline 1polarities
baseline 1representative citing papers
citing papers explorer
-
Quantization Meets Spikes: Nearly Lossless Conversion at the First Timestep via Polarity Multi-Spike Mapping
A signed quantized activation plus a multi-spike signed neuron makes a one-timestep spiking network reproduce the quantized ANN exactly, giving ANN-identical accuracy at T=1 on CIFAR-10/100 and state-of-the-art one-step accuracy on ImageNet and event datasets.