Nonlinear All-Optical Diffractive Deep Neural Network with 10.6 μm Wavelength for Image Classification

Table 9

The classification accuracy in the N-D²NN model adding the RReLU 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	93.10	94.93	95.08	95.81	96.06	96.15	96.35	96.78
	40	90.45	94.16	95.19	95.81	95.71	96.05	96.43	96.47
	50	85.01	93.78	95.12	95.39	95.90	95.99	96.28	96.23
	60	84.44	92.85	94.43	95.24	95.59	96.06	96.14	96.35
	70	68.88	91.27	93.71	95.03	95.49	95.67	95.85	96.08