Abstract:Abstract-In this paper, we report the development of a strictly non-blocking 8 × 8 silicon photonics switch designed to operate in the O-band. This 8 × 8 switch is based on path-independent insertion-loss topology and is composed of 2 × 2 thermo-optic double Mach-Zehnder switches and adiabatic intersections. The fabricated 8 × 8 switch chip is electrically packaged with a ceramic chip carrier and inserted into a socket on a printed circuit board. As for the optical connection, an optical fiber array and edge … Show more
“…The current training time is also limited by the current setting time (2-3 ms for one channel) of our current sources, limiting onchip training for a large dataset, which can also be improved if integrating drivers for heaters 34,35 . If the driver has a >100 MHz bandwidth, the training time will be dominated by the heater response, which is about tens of microseconds (μs) (~30 μs for our heaters, which can be shortened to <10 μs 34 ).…”
Section: Discussionmentioning
confidence: 99%
“…All experiments were performed in normal lab environment without any thermal management (Supplementary Section 5.3). One reason for the chip reproducibility is due to our advanced 12 inch silicon photonic platform (AIST-SCR) on which our large-scale photonic switches were fabricated with high reproducibility 34,35 . The other reason is that the PPC device for classification application presented in this work inherently has higher robustness to device imperfections than other devices for switching applications (see Supplementary Section 5.4).…”
On-chip training remains a challenging issue for photonic devices to implement machine learning algorithms. Most demonstrations only implement inference in photonics for offline-trained neural network models. On the other hand, artificial neural networks are one of the most deployed algorithms, while other machine learning algorithms such as supporting vector machine (SVM) remain unexplored in photonics. Here, inspired by SVM, we propose to implement projection-based classification principle by constructing nonlinear mapping functions in silicon photonic circuits and experimentally demonstrate on-chip bacterial foraging training for this principle to realize single Boolean logics, combinational Boolean logics, and Iris classification with ~96.7 − 98.3 per cent accuracy. This approach can offer comparable performances to artificial neural networks for various benchmarks even with smaller scales and without leveraging traditional activation functions, showing scalability advantage. Natural-intelligence-inspired bacterial foraging offers efficient and robust on-chip training, and this work paves a way for photonic circuits to perform nonlinear classification.
“…The current training time is also limited by the current setting time (2-3 ms for one channel) of our current sources, limiting onchip training for a large dataset, which can also be improved if integrating drivers for heaters 34,35 . If the driver has a >100 MHz bandwidth, the training time will be dominated by the heater response, which is about tens of microseconds (μs) (~30 μs for our heaters, which can be shortened to <10 μs 34 ).…”
Section: Discussionmentioning
confidence: 99%
“…All experiments were performed in normal lab environment without any thermal management (Supplementary Section 5.3). One reason for the chip reproducibility is due to our advanced 12 inch silicon photonic platform (AIST-SCR) on which our large-scale photonic switches were fabricated with high reproducibility 34,35 . The other reason is that the PPC device for classification application presented in this work inherently has higher robustness to device imperfections than other devices for switching applications (see Supplementary Section 5.4).…”
On-chip training remains a challenging issue for photonic devices to implement machine learning algorithms. Most demonstrations only implement inference in photonics for offline-trained neural network models. On the other hand, artificial neural networks are one of the most deployed algorithms, while other machine learning algorithms such as supporting vector machine (SVM) remain unexplored in photonics. Here, inspired by SVM, we propose to implement projection-based classification principle by constructing nonlinear mapping functions in silicon photonic circuits and experimentally demonstrate on-chip bacterial foraging training for this principle to realize single Boolean logics, combinational Boolean logics, and Iris classification with ~96.7 − 98.3 per cent accuracy. This approach can offer comparable performances to artificial neural networks for various benchmarks even with smaller scales and without leveraging traditional activation functions, showing scalability advantage. Natural-intelligence-inspired bacterial foraging offers efficient and robust on-chip training, and this work paves a way for photonic circuits to perform nonlinear classification.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.