Modeling Silicon-Photonic Neural Networks under Uncertainties
This work addresses a critical reliability and performance problem for designers and users of silicon-photonic neural networks, which are touted for high-speed and energy-efficient computing.
This paper investigates the impact of random uncertainties from fabrication and thermal variations on the classification accuracy of Mach-Zehnder Interferometer (MZI)-based silicon-photonic neural networks (SPNNs). The study reveals that these uncertainties can cause a catastrophic 70% accuracy loss in an SPNN with two hidden layers and 1374 tunable-thermal-phase shifters, even with mature fabrication processes.
Silicon-photonic neural networks (SPNNs) offer substantial improvements in computing speed and energy efficiency compared to their digital electronic counterparts. However, the energy efficiency and accuracy of SPNNs are highly impacted by uncertainties that arise from fabrication-process and thermal variations. In this paper, we present the first comprehensive and hierarchical study on the impact of random uncertainties on the classification accuracy of a Mach-Zehnder Interferometer (MZI)-based SPNN. We show that such impact can vary based on both the location and characteristics (e.g., tuned phase angles) of a non-ideal silicon-photonic device. Simulation results show that in an SPNN with two hidden layers and 1374 tunable-thermal-phase shifters, random uncertainties even in mature fabrication processes can lead to a catastrophic 70% accuracy loss.