2019/05/31 by Rubab Amin, Amin, Rubab, Jonathan George +19 · 1 citation
Computer Science · Engineering · Physics and Astronomy · #Applied Physics (physics.app-ph) #FOS: Physical sciences #Mechanical and Optical Resonators #Neural Networks and Reservoir Computing #Optics (physics.optics) #Photonic and Optical Devices
paper · pdf · doi:10.48550/arxiv.1906.00017
openalex publication_date 2019/05/31 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28
Recently integrated optics has become an intriguing platform for implementing\nmachine learning algorithms and inparticular neural networks. Integrated\nphotonic circuits can straightforwardly perform vector-matrix\nmultiplicationswith high efficiency and low power consumption by using\nweighting mechanism through linear optics. Although,this can not be said for\nthe activation function which requires either nonlinear optics or an\nelectro-optic module withan appropriate dynamic range. Even though all-optical\nnonlinear optics is potentially faster, its current integrationis challenging\nand is rather inefficient. Here we demonstrate an electro-absorption modulator\nbased on an IndiumTin Oxide layer, whose dynamic range is used as nonlinear\nactivation function of a photonic neuron. The nonlinearactivation mechanism is\nbased on a photodiode, which integrates the weighed products, and whose\nphotovoltage drivesthe elecro-absorption modulator. The synapse and neuron\ncircuit is then constructed to execute a 200-node MNISTclassification neural\nnetwork used for benchmarking the nonlinear activation function and compared\nwith an equivalentelectronic module.\n