Fig. 5: Temporally programmable holographic neural network.
a Schematic of two-layer holographic neural network inference with sequential handwritten digit and handwritten digit + fashion product classification functionalities at times T1 and T2. Detectors 0–2 correspond to digits 0–2, while detectors 3, 4, and 5 correspond to t-shirts, trousers, and shoes, respectively. b Schematic of the optical interconnections between the two holographic layers at times T1 and T2. c, d The calculated and experimental amplitude hologram profiles and corresponding classification results. Scale bars, 100 and 3 μm. Representative input images, camera images with red boxes indicating the targeted detector regions, the energy distribution percentage in the targeted regions, and the confusion matrix are presented.
