Design of optical neural networks with component imprecisions

Abstract

For the benefit of designing scalable, fault resistant optical neural networks (ONNs), we investigate the effects architectural designs have on the ONNs' robustness to imprecise components. We train two ONNs -- one with a more tunable design (GridNet) and one with better fault tolerance (FFTNet) -- to classify handwritten digits. When simulated without any imperfections, GridNet yields a better accuracy (~98%) than FFTNet (~95%). However, under a small amount of error in their photonic components, the more fault tolerant FFTNet overtakes GridNet. We further provide thorough quantitative and qualitative analyses of ONNs' sensitivity to varying levels and types of imprecisions. Our results offer guidelines for the principled design of fault-tolerant ONNs as well as a foundation for further research.

References in corpus (10)

• Deep Learning with Coherent Nanophotonic Circuits
• Universal Linear Optics
• Neuromorphic Silicon Photonic Networks
• Optoelectronic Reservoir Computing
• Quantum transport simulations in a programmable nanophotonic processor
• Thermally-Reconfigurable Quantum Photonic Circuits at Telecom Wavelength by Femtosecond Laser Micromachining
• 60dB high-extinction auto-configured Mach--Zehnder interferometer
• Direct dialling of Haar random unitary matrices
• Benchmarking integrated photonic architectures
• Analog Optical Computing Primitives in Silicon Photonics

Cited by in corpus (17)

• Hardware error correction for programmable photonics
• Efficient training and design of photonic neural network through neuroevolution
• Asymptotically Fault-Tolerant Programmable Photonics
• Inverse-design of high-dimensional quantum optical circuits in a complex medium
• Accurate Self-Configuration of Rectangular Multiport Interferometers
• Photonic convolutional neural networks using integrated diffractive optics
• Stability of Self-Configuring Large Multiport Interferometers
• Scalable and Robust Photonic Integrated Unitary Converter
• Characterization and Optimization of Integrated Silicon-Photonic Neural Networks under Fabrication-Process Variations
• Characterizing Coherent Integrated Photonic Neural Networks under Imperfections
• Multiplexed gradient descent: Fast online training of modern datasets on hardware neural networks without backpropagation
• On-chip Optical Phase Monitoring in Multi-Transverse-Mode Integrated Silicon-based Optical Processors
• Bayesian Photonic Accelerators for Energy Efficient and Noise Robust Neural Processing
• Quantum Optical Convolutional Neural Network: A Novel Image Recognition Framework for Quantum Computing
• The Impact of Feature Representation on the Accuracy of Photonic Neural Networks
• Sub-quadratic scalable approximate linear converter using multi-plane light conversion with low-entropy mode mixers
• Emergence of biased errors in imperfect photonic circuits