On-chip photonic diffractive optical neural network based on a spatial domain electromagnetic propagation model

Fu, Tingzhao; Zang, Yubin; Huang, Honghao; Du, Z.; Hu, Chengyang; Chen, Minghua; Yang, Sigang; Chen, Hongwei

doi:10.1364/oe.435183

Cited by 37 publications

(50 citation statements)

References 33 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…1a, b ). Both amplitude modulation and the phase shift of the transmitted wave can be programmed by adjusting the width and length of the subwavelength slots, respectively 47 , 48 (Supplementary Fig. S2 ).…”

Section: Resultsmentioning

confidence: 99%

Integrated photonic metasystem for image classifications at telecommunication wavelength

et al. 2022

View full text Add to dashboard Cite

Miniaturized image classifiers are potential for revolutionizing their applications in optical communication, autonomous vehicles, and healthcare. With subwavelength structure enabled directional diffraction and dispersion engineering, the light propagation through multi-layer metasurfaces achieves wavelength-selective image recognitions on a silicon photonic platform at telecommunication wavelength. The metasystems implement high-throughput vector-by-matrix multiplications, enabled by near 103 nanoscale phase shifters as weight elements within 0.135 mm2 footprints. The diffraction manifested computing capability incorporates the fabrication and measurement related phase fluctuations, and thus the pre-trained metasystem can handle uncertainties in inputs without post-tuning. Here we demonstrate three functional metasystems: a 15-pixel spatial pattern classifier that reaches near 90% accuracy with femtosecond inputs, a multi-channel wavelength demultiplexer, and a hyperspectral image classifier. The diffractive metasystem provides an alternative machine learning architecture for photonic integrated circuits, with densely integrated phase shifters, spatially multiplexed throughput, and data processing capabilities.

show abstract

Section: Resultsmentioning

confidence: 99%

Integrated photonic metasystem for image classifications at telecommunication wavelength

et al. 2022

View full text Add to dashboard Cite

show abstract

“…One important issue in an optical neural network design is its final experimental inference capability. Most of the diffractive optical neural networks proposed up to now, show high percentage of consistency between numerical predictions and experimental verifications [26][27][28][29][30]. For our design, however, the matching between numerical testing results and experimental testing results should be 100% for accurate and precise performance of multi-functional logic gate.…”

Section: Design Considerationsmentioning

confidence: 86%

“…Locally periodic approximation that assumes the metasurface is locally periodic (i.e., is periodic over any small region) can be a high-speed alternative that calculate the field across the plane right after the metasurface, by using small full-wave simulations to compute the field transmission phase and amplitude for each scattering element (meta-atom). Thereafter, by using near-to-far field transformation [22][23], or scalar-wave approximation [24][25], or spatial domain electromagnetic propagation [26], the fields can be propagated between metasurface layers. In this article, near-to-far field transformation [22][23] is utilized to propagate the fields.…”

Section: Modeling A) Forward Propagationmentioning

confidence: 99%

Realization of optical logic gates using on-chip diffractive optical neural networks

Zarei

Khavasi

2022

Preprint

View full text Add to dashboard Cite

Optical computing is highly desired as a potential strategy for circumventing the performance limitations of semiconductor-based electronic devices and circuits. Optical logic gates are considered as fundamental building blocks for optical computation and they enable logic functions to be performed extremely quickly without the generation of heat and crosstalk. Here, we discuss the design of a multi-functional optical logic gate based on an on-chip diffractive optical neural network that can perform AND, NOT and OR logic operations at the wavelength of 1.55µm. The wavelength-independent operation of the multi-functional logic gate at seven wavelengths (over a bandwidth of 60nm) is also studied which paves the way for wavelength division multiplexed parallel computation. This simple, highly-integrable, low-loss, energy-efficient and broadband optical logic gate provides a path for the development of high-speed on-chip nanophotonic processors for future optical computing applications.

show abstract

“…In the future, the free-space optical components can be replaced by the on-chip lenses which have been achieved in the nanophotonic convolver [39], on-chip deep learning systems [40][41][42][43] and integrated tunable varifocal lenses performing like error backpropagation in neural networks [44]. Once combined with on-chip lenses, our nonlinear JTC will have high energy efficiency, compact volume and high speed, and therefore promotes the potential convolution-related applications in many aspects, including image classification with large and deep convolutional neural networks, speech recognition and translation with the combination of convolutional neural networks and deep recurrent neural networks, autonomous driving by mapping raw pixels with convolutional neural networks, inverse design (like nanophotonic / plasmonic structures, self-adaptive microwave cloak) problem solving based on convolutional neural networks, and robotic manipulation with deep reinforcement learning.…”

Section: Discussionmentioning

confidence: 99%

Nonlinear Optical Joint Transform Correlator for Low Latency Convolution Operations

George¹,

Solyanik‐Gorgone²,

Yang³

et al. 2022

Preprint

View full text Add to dashboard Cite

Convolutions are one of the most relevant operations in artificial intelligence (AI) systems. High computational complexity scaling poses significant challenges, especially in fast-responding networkedge AI applications. Fortunately, the Convolution Theorem can be executed 'on-the-fly' in the optical domain via a joint transform correlator (JTC) offering to fundamentally reduce the computational complexity. Nonetheless, the iterative two-step process of a classical JTC renders them unpractical. Here we introduce a novel implementation of an optical convolution-processor capable of near-zero latency by utilizing all-optical nonlinearity inside a JTC, thus minimizing electronic signal or conversion delay. Fundamentally we show how this nonlinear auto-correlator enables reducing the high O(n 4 ) scaling complexity of processing two-dimensional data to only O(n 2 ). Moreover, this optical JTC processes millions of channels in time-parallel, ideal for large-matrix machine learning tasks. Exemplary utilizing the nonlinear process of four-wave mixing, we show light processing performing a full convolution that is temporally limited only by geometric features of the lens and the nonlinear material's response time. We further discuss that the all-optical nonlinearity exhibits gain in excess of > 10 3 when enhanced by slow-light effects such as epsilon-near-zero. Such novel implementation for a machine learning accelerator featuring low-latency and non-iterative massive data parallelism enabled by fundamental reduced complexity scaling bears significant promise for network-edge, and cloud AI systems.

show abstract

On-chip photonic diffractive optical neural network based on a spatial domain electromagnetic propagation model

Cited by 37 publications

References 33 publications

Integrated photonic metasystem for image classifications at telecommunication wavelength

Integrated photonic metasystem for image classifications at telecommunication wavelength

Realization of optical logic gates using on-chip diffractive optical neural networks

Nonlinear Optical Joint Transform Correlator for Low Latency Convolution Operations

Contact Info

Product

Resources

About