Classification on CIFAR10-DVS
85.9Accuracysliding PSN (k=2)
Evaluation results
| Method | Links | ||
|---|---|---|---|
| sliding PSN (k=2)Spiking Network=VGG, Time-steps=102023.04 | 85.9 | — | |
| EventRPG (CAM)Training Method=STBP, Neural Network=Spike-VGG11, Neuron=LIF, Timesteps=10, Resolution=(48, 48)2024.03 | 85.55 | — | |
| FPTArchitecture=VGG-11, Parallel=true, Serial=false, T=102026.02 | 85.5 | — | |
| sliding PSN (k=2)Spiking Network=VGG, Time-steps=82023.04 | 85.3 | — | |
| sliding PSNArchitecture=VGGSNN, Parallel=true, Serial=true, T=82026.02 | 85.3 | — | |
| DSNArchitecture=VGGSNN, Parallel=true, Serial=true, T=82026.02 | 85.3 | — | |
| EventRPG (Saliency Map)Training Method=STBP, Neural Network=Spike-VGG11, Neuron=LIF, Timesteps=10, Resolution=(48, 48)2024.03 | 84.96 | — | |
| TEBNSpiking Network=VGG, Time-steps=102023.04 | 84.9 | — | |
| SpikingResformerType=Transfer Learning, Architecture=SpikingResformer-S, #Param (M)=17.25, T=102024.03 | 84.8 | — | |
| SpikingResformerType=Transfer Learning, Architecture=SpikingResformer-Ti, #Param (M)=10.76, T=102024.03 | 84.7 | — | |
| DSNArchitecture=VGGSNN, Parallel=true, Serial=true, T=42026.02 | 83.9 | — | |
| MorphSNNArchitecture=DGD-SNN-7-3, T=162026.03 | 83.7 | — | |
| TETT=102022.10 | 83.32 | — | |
| TCJAT=10, Protocol=TET2022.10 | 83.3 | — | |
| Evolutionary TrainingType=SNN training, Architecture=ResNet18, Timestep=102024.01 | 83.3 | — | |
| VGGSNNT=102026.01 | 83.2 | — | |
| TETSpiking Network=VGG, Time-steps=102023.04 | 83.17 | — | |
| FSTA-SNNT=162026.04 | 82.7 | — | |
| sliding PSN (k=2)Spiking Network=VGG, Time-steps=42023.04 | 82.3 | — | |
| sliding PSNArchitecture=VGGSNN, Parallel=true, Serial=true, T=42026.02 | 82.3 | — | |
| CP-DSAArchitecture=ResNet-14, Timestep=82025.12 | 82.2 | — | |
| Vision SmolMambaT=162026.04 | 82.2 | — | |
| Shortcut Back-propagationType=SNN training, Architecture=ResNet18, Timestep=102024.01 | 82 | — | |
| RPSUArchitecture=VGGSNN, Parallel=true, Serial=false, T=102026.02 | 82 | — | |
| CP-DSAArchitecture=ResNet-14, Timestep=52025.12 | 81.8 | — | |
| tdBN (w/ NDA)Model=VGG-11, T Step=10, Resolution=128x1282022.03 | 81.7 | — | |
| NDATraining Method=STBP-tdBN [Zheng et al., 2021], Neural Network=Spike-VGG11, Neuron=LIF, Timesteps=10, Resolution=(128,128)2024.03 | 81.7 | — | |
| CogniSNNArchitecture=ER-RGA-7, T=162026.03 | 81.6 | — | |
| QB-LIFArchitecture=ResNet-20, Time Step=102026.04 | 81.6 | — | |
| SpikingResformerType=Direct Training, Architecture=SpikingResformer-Ti*, #Param (M)=10.79, T=102024.03 | 81.5 | — | |
| EventmixTraining Method=STBP, Neural Network=Pre-Act Resnet18, Neuron=PLIF, Timesteps=10, Resolution=(48,48)2024.03 | 81.45 | — | |
| STSCT=10, Protocol=MSE2022.10 | 81.4 | — | |
| SpikingformerType=Direct Training, Architecture=Spikingformer-2-256, #Param (M)=2.55, T=162024.03 | 81.3 | — | |
| SpikingformerArchitecture=Spikingformer-2-256, T=162026.03 | 81.3 | — | |
| DeepTAGEArchitecture=VGG-11, Parallel=false, Serial=true, T=102026.02 | 81.23 | — | |
| CTSNArchitecture=ResNet20, T=102026.01 | 81.23 | — | |
| CP-DSAArchitecture=ResNet-14, Timestep=102025.12 | 81.2 | — | |
| SpikformerSpike-driven=false, T=162023.07 | 80.9 | — | |
| SpikformerArchitecture=Spikformer-2-256, T=162026.03 | 80.9 | — | |
| SpikformerT=162026.04 | 80.9 | — | |
| TCJAT=10, Protocol=MSE2022.10 | 80.7 | — | |
| TCJA-SNNT=102026.04 | 80.7 | — | |
| SpikformerType=Direct Training, Architecture=Spikformer-2-256, #Param (M)=2.55, T=162024.03 | 80.6 | — | |
| IMLIFArchitecture=VGG-13, Time Step=402026.04 | 80.5 | — | |
| Ternary SpikeArchitecture=ResNet20, T=102026.01 | 80.3 | — | |
| Spike-driven TransformerSpike-driven=true, T=162023.07 | 80 | — | |
| Spike-driven TransformerType=Direct Training, Architecture=Spike-driven Transformer-2-256, #Param (M)=2.55, T=162024.03 | 80 | — | |
| Spike-driven TransformerArchitecture=Spiking Transformer, Timestep=162025.12 | 80 | — | |
| Spike-D- TransformerArchitecture=S-d-Transformer-2-256, T=162026.03 | 80 | — | |
| SDTT=162026.04 | 80 | — | |
| SpikingformerType=Direct Training, Architecture=Spikingformer-2-256, #Param (M)=2.55, T=102024.03 | 79.9 | — | |
| MorphSNNArchitecture=DGD-SNN-7-3, T=52026.03 | 79.9 | — | |
| Ternary SpikeT=102026.04 | 79.8 | — | |
| Ternary SpikeArchitecture=ResNet-20, Time Step=102026.04 | 79.8 | — | |
| tdBN (w/ NDA)Model=VGG-11, T Step=102022.03 | 79.6 | — | |
| NDATraining Method=STBP-tdBN [Zheng et al., 2021], Neural Network=Spike-VGG11, Neuron=LIF, Timesteps=10, Resolution=(48,48)2024.03 | 79.6 | — | |
| LM-HArchitecture=ResNet-19, Timestep=102025.12 | 79.1 | — | |
| SparseSpikformerT=162026.04 | 79.1 | — | |
| CTSNArchitecture=VGG16, T=102026.01 | 79.06 | — | |
| CLIFArchitecture=VGG11, T=162026.01 | 79 | — | |
| CogniSNNArchitecture=WS-RGA-7, T=52026.03 | 79 | — | |
| IdentityTraining Method=TET [Deng et al., 2022]2024.03 | 78.85 | — | |
| SpikingReformerArchitecture=SpikingReformer-4-384, T=162026.03 | 78.8 | — | |
| MPBNArchitecture=ResNet-20, Time Step=102026.04 | 78.7 | — | |
| SpikformerType=Direct Training, Architecture=Spikformer-2-256, #Param (M)=2.55, T=102024.03 | 78.6 | — | |
| SSNNArchitecture=VGG-9, Timestep=82025.12 | 78.57 | — | |
| GLIFSpiking Network=Wide 7B Net, Time-steps=162023.04 | 78.1 | — | |
| GLIFArchitecture=Wide 7B Net, Parallel=false, Serial=true, T=162026.02 | 78.1 | — | |
| GLIFArchitecture=7B-wideNet, T=162026.01 | 78.1 | — | |
| GLIFArchitecture=7B-wideNet, Time Step=162026.04 | 78.1 | — | |
| Spikformer w/ CPG-PESNN=true, Spike PE=true, Param (M)=2.062024.05 | 78.06 | — | |
| tdBN (w/ NDA)Model=ResNet-19, T Step=10, Quadrupled channel number=true2022.03 | 78 | — | |
| Spikformer w/ CPG-PE [Equal Param]SNN=true, Spike PE=true, Param (M)=1.992024.05 | 78 | — | |
| Spikformer w/ RPESNN=true, Spike PE=true, Param (M)=2.572024.05 | 77.95 | — | |
| EventDropTraining Method=TET [Deng et al., 2022]2024.03 | 77.73 | — | |
| Spikformer w/ Float-PESNN=true, Spike PE=false, Param (M)=1.992024.05 | 77.6 | — | |
| SpikingformerArchitecture=Spikingformer-2-256, T=52026.03 | 77.5 | — | |
| DSRT=102022.10 | 77.41 | — | |
| STL-SNNMethod=Spike-based BP, Selection=Best top-12022.06 | 77.3 | — | |
| DSRSpike-driven=true, T=102023.07 | 77.3 | — | |
| DSRT=102026.04 | 77.3 | — | |
| OTTT_ONetwork structure=VGG (SWS), Params=9.2M, Time steps=102022.10 | 77.1 | — | |
| Spikformer w/ Random-PESNN=true, Spike PE=true, Param (M)=2.062024.05 | 76.44 | — | |
| Spikformer w/o PESNN=true, Spike PE=false, Param (M)=1.992024.05 | 76.4 | — | |
| tdBN (w/o NDA)Model=VGG-11, T Step=10, Resolution=128x128, Our implementation=true2022.03 | 76.3 | — | |
| OTTT_ANetwork structure=VGG (SWS), Params=9.2M, Time steps=102022.10 | 76.3 | — | |
| tdBN (w/o NDA)Model=VGG-11, T Step=10, Our implementation=true2022.03 | 76.2 | — | |
| DspikeSpike-driven=false, T=102023.07 | 75.4 | — | |
| DspikeSpiking Network=ResNet-18, Time-steps=102023.04 | 75.4 | — | |
| DspikeType=Direct Training, Architecture=ResNet-18, #Param (M)=11.21, T=102024.03 | 75.4 | — | |
| DspikeType=SNN training, Architecture=ResNet18, Timestep=102024.01 | 75.4 | — | |
| DspikeT=102026.04 | 75.4 | — | |
| PLIFT=202022.10 | 74.8 | — | |
| PLIFMethod=Spike-based BP2022.06 | 74.8 | — | |
| PLIFModel=VGG-11, T Step=20, Resolution=128x1282022.03 | 74.8 | — | |
| BPTTNetwork structure=7-layer CNN (PLIF, BN), Params=1.1M, Time steps=202022.10 | 74.8 | — | |
| PLIFSpike-driven=true, T=202023.07 | 74.8 | — | |
| PLIFSpiking Network=PLIF Net, Time-steps=202023.04 | 74.8 | — | |
| PLIFType=Direct Training, Architecture=5 Conv, 2 FC, #Param (M)=17.22, T=202024.03 | 74.8 | — | |
| PLIFArchitecture=PLIF Net, T=202026.01 | 74.8 | — |