The new concepts and architectures are still frequently tested on MNIST. However, we argue that the MNIST data set does not include temporal information and does not provide spike events generated from sensors. Compared to a static data set, a dynamic data set contains richer temporal features and therefore is more suitable to exploit an SNN's potential ability. The event-based benchmark data sets include N-MNIST (Orchard, Jayawant, Cohen, & Thakor, 2015), CIFAR10-DVS (Hongmin Li, Liu, Ji, Li, & Shi, 2017), N-CARS (Sironi, Brambilla, Bourdis, Lagorce, & Benosman, 2018), DVS-Gesture (Amir et al., 2017), and SHD (Cramer, Stradmann, Schemmel, & Zenke, 2020). Table 3 shows the models for developing SNNs—their architectures and learning type along with their accuracy rates on the neuromorphic data sets.
Summary of Recent SNN Learning Models and Their Accuracy on Event-Based Data Set.
| Reference . | Network Type . | Learning Rule and Structure Configuration . | Data Set . | CA % . |
|---|---|---|---|---|
| Kugele et al. (2020) | SNN | ANN-to-SNN conversion | N-MNIST | 95.54 |
| CIFAR-DVS | 66.61 | |||
| DvsGesture | 96.97 | |||
| N-Cars | 94.07 | |||
| Wu et al. (2018) | Spiking MLP | Spatiotemporal backpropagation (STBP) 34 34 2-800-10 | N_MNIST | 98.78 |
| Wu et al. (2019) | SNN | Spatiotemporal backpropagation (STBP) 128C3(Encoding)-128C3-AP2-384C3-384C3-AP2-1024FC-512FC-Voting | N-MNIST | 99.53 |
| CIFAR-DVS | 60.5 | |||
| Zheng et al. (2020) | ResNet17 SNN | Threshold-dependent batch normalization method based on spatiotemporal backpropagation (STBP-tdBN) | CIFAR-DVS | 67.80 |
| DvsGesture | 96.87 | |||
| Lee et al. (2016) | SNN | Supervised backpropagation (34 34 2)-800-10 | N-MNIST | 98.66 |
| Yao et al. (2021) | Spiking CNN | Temporal-wise attention SNN (TA-SNN) | DvsGesture | 98.61 |
| (1) Input-MP4-64C3-128C3-AP2-128C3-AP2-256FC-11 | CIFAR-DVS | 72 | ||
| (2) Input-32C3-AP2-64C3-AP2-128C3-AP2-256C3-AP2-512C3-AP4-256FC-10 | SHD | 91.08 | ||
| (3) Input-128FC-128FC-20 | ||||
| Neil and Liu (2016) | Spiking CNN | ANN-to-SNN conversion | N-MNIST | 95.72 |
| Reference . | Network Type . | Learning Rule and Structure Configuration . | Data Set . | CA % . |
|---|---|---|---|---|
| Kugele et al. (2020) | SNN | ANN-to-SNN conversion | N-MNIST | 95.54 |
| CIFAR-DVS | 66.61 | |||
| DvsGesture | 96.97 | |||
| N-Cars | 94.07 | |||
| Wu et al. (2018) | Spiking MLP | Spatiotemporal backpropagation (STBP) 34 34 2-800-10 | N_MNIST | 98.78 |
| Wu et al. (2019) | SNN | Spatiotemporal backpropagation (STBP) 128C3(Encoding)-128C3-AP2-384C3-384C3-AP2-1024FC-512FC-Voting | N-MNIST | 99.53 |
| CIFAR-DVS | 60.5 | |||
| Zheng et al. (2020) | ResNet17 SNN | Threshold-dependent batch normalization method based on spatiotemporal backpropagation (STBP-tdBN) | CIFAR-DVS | 67.80 |
| DvsGesture | 96.87 | |||
| Lee et al. (2016) | SNN | Supervised backpropagation (34 34 2)-800-10 | N-MNIST | 98.66 |
| Yao et al. (2021) | Spiking CNN | Temporal-wise attention SNN (TA-SNN) | DvsGesture | 98.61 |
| (1) Input-MP4-64C3-128C3-AP2-128C3-AP2-256FC-11 | CIFAR-DVS | 72 | ||
| (2) Input-32C3-AP2-64C3-AP2-128C3-AP2-256C3-AP2-512C3-AP4-256FC-10 | SHD | 91.08 | ||
| (3) Input-128FC-128FC-20 | ||||
| Neil and Liu (2016) | Spiking CNN | ANN-to-SNN conversion | N-MNIST | 95.72 |