Deep Neural Network Through an InP SOA-Based Photonic Integrated Cross-Connect

• Published in: IEEE journal of selected topics in quantum electronics Vol. 26; no. 1; pp. 1 - 11
• Main Authors: Shi, Bin, Calabretta, Nicola, Stabile, Ripalta
• Format: Journal Article
• Language: English
• Published: New York IEEE 01.01.2020; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Accuracy; Arrayed waveguide gratings; Artificial neural networks; Biological neural networks; Computation; Error analysis; Feedback loops; image classification; Indium phosphide; Linear functions; Neural networks; Neurons; Optimization; Phosphides; photonic integrated circuits; Photonics; Power consumption; Semiconductor optical amplifiers; Semiconductors; Wavelength division multiplexing
• ISSN: 1077-260X; 1558-4542
• DOI: 10.1109/JSTQE.2019.2945548

Photonic neuromorphic computing is raising a growing interest as it promises to provide massive parallelism and low power consumption. In this paper, we demonstrate for the first time a feed-forward neural network via an 8 × 8 Indium Phosphide cross-connect chip, where up to 8 on-chip weighted addition circuits are co-integrated, based on semiconductor optical amplifier technology. We perform the weight calibration per neuron, resulting in a normalized root mean square error smaller than 0.08 and a best case dynamic range of 27 dB. The 4 input to 1 output weighted addition operation is executed on-chip and is part of a neuron, whose non-linear function is implemented via software. A three feedback loop optimization procedure is demonstrated to enable an output neuron accuracy improvement of up to 55%. The exploitation of this technology as neural network is evaluated by implementing a trained 3-layer photonic deep neural network to solve the Iris flower classification problem. Prediction accuracy of 85.8% is achieved, with respect to the 95% accuracy obtained via a computer. A comprehensive analysis of the error evolution in our system reveals that the electrical/optical conversions dominate the error contribution, which suggests that an all optical approach is preferable for future neuromorphic computing hardware design.