Table 4.
Performance comparison between the proposed method and the state-of-the-art methods on different datasets.
| Method | SHD | N-MNIST | CIFAR10-DVS | DVS128 Gesture | ||||
|---|---|---|---|---|---|---|---|---|
| T | Acc. (%) | T | Acc. (%) | T | Acc. (%) | T | Acc. (%) | |
| Slayer (Shrestha and Orchard, 2018) | - | - | 300 | 99.22 | - | - | 1,600 | 93.4 |
| HATS (Sironi et al., 2018) | - | - | - | 99.10 | - | 52.4 | - | - |
| DART (Ramesh et al., 2019) | - | - | - | 97.95 | - | 65.8 | - | - |
| NeuNorm (Wu et al., 2019) | - | - | - | 99.53 | - | 60.5 | - | - |
| Rollout (Kugele et al., 2020) | - | - | 32 | 99.57 | 48 | 66.97 | 240 | 97.27 (10 classes) |
| DECOLLE (Kaiser et al., 2020) | - | - | - | - | - | - | 500 | 95.7 |
| LISNN (Cheng et al., 2020) | - | - | 20 | 99.45 | - | - | - | - |
| tdBN (Zheng et al., 2021) | - | - | - | - | 10 | 67.8 | 40 | 96.87 |
| LIAF-Net (Wu et al., 2021) | - | - | - | 10 | 70.40 | 60 | 97.56 | |
| PLIF (Fang et al., 2021c) | - | - | 10 | 99.61 | 20 | 74.80 | 20 | 97.57 |
| LIF RSNN (Cramer et al., 2020) | 2,000 | 73.3 | - | - | - | - | - | - |
| Hetero. RSNN (Perez-Nieves et al., 2021) | - | 83.5 | - | 97.5 | - | - | - | 82.9 |
| RELU SRNN (Yin et al., 2020) | 250 | 88.93 | - | - | - | - | - | - |
| Adaptive SRNN (Yin et al., 2021) | 250 | 90.4 | - | - | - | - | - | - |
| SEW-ResNet (Fang et al., 2021b) | - | - | - | - | 16 | 74.4 | 16 | 97.92 |
| TA-SNN (Yao et al., 2021) | 15 | 91.08 | - | - | 10 | 72.00 | 20 | 98.61 |
| Dspike (Li et al., 2021) | - | - | - | - | 10 | 75.45 | - | - |
| SALT (Kim and Panda, 2021) | - | - | - | 20 | 67.1 | - | - | |
| TET (Deng et al., 2022) | - | - | - | - | 10 | 83.32 | - | - |
| DSR (Meng et al., 2022) | - | - | - | - | 10 | 77.41 | - | - |
| TCJA (Zhu et al., 2022) | - | - | - | - | 10 | 80.7(MSE) 83.3(TET) |
20 | 99.0 |
| STSC (this work) | 15 | 92.36 | 10 | 99.64 | 10 | 81.4(MSE) | 20 | 98.96 |
The bold values indicate the best outcomes for each dataset.