Skip to main content
. Author manuscript; available in PMC: 2021 Jan 1.
Published in final edited form as: IEEE J Sel Top Quantum Electron. 2019 Jun 6;26(1):3700114. doi: 10.1109/JSTQE.2019.2921376

Fig. 2.

Fig. 2.

Convergence plots and confusion matrices for all-optical D2NN-based classification of handwritten digits (MNIST dataset). (a) Convergence curve and confusion matrix for a phase-only, fully-connected D2NN (ΔZ=40λ) design. (b) Convergence curve and confusion matrix for a phase-only, partially-connected D2NN (ΔZ=4λ) design. (c) and (d) are counterparts of (a) and (b), respectively, for complex-modulation D2NN designs, where both the amplitude and phase of each neuron are trainable parameters.