Image Classification on MNIST (Quantization Accuracy Metrics)
98.67Accuracy (W32A32)DoReFa
Evaluation results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| DoReFaArchitecture=KAGN Simple2025.11 | 98.67 | 98.7 | 98.71 | 98.58 | 98.33 | 97.93 | 97.96 | 96.22 | |
| DSQArchitecture=KAGN Simple2025.11 | 98.67 | 98.43 | 98.24 | 98.64 | 85.14 | 96.66 | 93.28 | 85.14 | |
| LSQArchitecture=KAGN Simple2025.11 | 98.67 | 98.95 | 98.69 | 98.74 | 98.34 | 86.25 | 78.42 | 76.32 | |
| DSQArchitecture=KAN FCN2025.11 | 98 | 98.43 | 98.24 | 98.13 | 97.71 | 97.71 | 97.82 | 97.71 | |
| LSQ+Architecture=KAN FCN2025.11 | 98 | 98.24 | 98.13 | 98.03 | 97.79 | 97.87 | 97.73 | 96.97 | |
| PACTArchitecture=KAN FCN2025.11 | 98 | 98.17 | 98.14 | 98.07 | 97.21 | 97.77 | 97.64 | 96.09 | |
| DoReFaArchitecture=KAN ConvNet2025.11 | 95.7 | 95.41 | 95.67 | 95.17 | 93.55 | 91.09 | 91.08 | 88.52 | |
| DSQArchitecture=KAN ConvNet2025.11 | 95.7 | 95.81 | 95.44 | 95.62 | 89.54 | 90.4 | 90.14 | 89.54 | |
| LSQ+Architecture=KAN ConvNet2025.11 | 95.7 | 95.7 | 95.56 | 95.17 | 94.46 | 90.95 | 94.11 | 92.04 |