Research Article
Nonlinear All-Optical Diffractive Deep Neural Network with 10.6 μm Wavelength for Image Classification
Table 7
The classification accuracy in the N-D2NN model adding the Leaky-ReLU activation function corresponds to pixel size and diffraction grating spacing.
| Accuracy (%) | Pixel size | 30 × 30 | 40 × 40 | 50 × 50 | 60 × 60 | 70 × 70 | 80 × 80 | 90 × 90 | 100 × 100 |
| Spacing (λ) | 30 | 93.16 | 94.98 | 95.20 | 95.83 | 96.30 | 96.01 | 96.51 | 96.58 | 40 | 90.40 | 94.13 | 95.39 | 95.86 | 95.95 | 95.96 | 96.35 | 96.55 | 50 | 80.79 | 93.74 | 95.04 | 95.64 | 95.79 | 95.98 | 96.35 | 96.55 | 60 | 77.74 | 92.52 | 94.00 | 95.51 | 95.63 | 95.83 | 96.21 | 96.44 | 70 | 68.84 | 89.87 | 93.73 | 95.16 | 95.42 | 96.03 | 96.10 | 96.25 |
|
|