Runtime Efficiency and Accuracy on MNIST (Image Classification)
98.24AccuracyQDSNN
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| QDSNNArchitecture=LeNet-5, Thresholds=0.6/0.72026.03 | 98.24 | 162 | 0.004 | 0.58 | |
| QDSNNArchitecture=5-Layer MLP, Thresholds=0.6/0.72026.03 | 97.8 | 83 | 0.012 | 0.99 | |
| SNNArchitecture=LeNet-5, Timesteps (T)=322026.03 | 97.76 | 552 | 0.053 | 29.26 | |
| QDSNNArchitecture=LeNet-5, Thresholds=0.2/0.42026.03 | 96.01 | 159 | 0.004 | 0.58 | |
| QDSNNArchitecture=5-Layer MLP, Thresholds=0.2/0.42026.03 | 95.5 | 105 | 0.009 | 0.95 | |
| SNNArchitecture=5-Layer MLP, Timesteps (T)=322026.03 | 95 | 539 | 0.08 | 43.1 | |
| QSNNArchitecture=5-Layer MLP, Timesteps (T)=42026.03 | 94.5 | 420 | 0.005 | 2.1 | |
| QSNNArchitecture=LeNet-5, Timesteps (T)=42026.03 | 93.09 | 70 | 0.001 | 0.07 |