On-chip CMOS-compatible optical signal processor

Yang, Lin; Ji, Ruiqiang; Zhang, Lei; Ding, Jianfeng; Xu, Qianfan

doi:10.1364/oe.20.013560

Cited by 79 publications

(44 citation statements)

References 22 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In this work, we propose an all-photonic SNN computing primitive, based on GST-based photonic neural elements, which attempts to bridge the gap between devices to system-level implementation of Photonic neural networks. We leverage the inherent wavelength division multiplexing (WDM) [27] property of optical networks to propose a non-volatile synaptic array, while exploring and mitigating the challenges arising from designs based on ring resonators of radii comparable to the wavelength of operation. Such a synaptic array can achieve higher densities compared to current state-of-art photonic computing systems.…”

Section: Introductionmentioning

confidence: 99%

Photonic In-Memory Computing Primitive for Spiking Neural Networks Using Phase-Change Materials

2019

View full text Add to dashboard Cite

Spiking Neural Networks (SNNs) offer an event-driven and more biologically realistic alternative to standard Artificial Neural Networks based on analog information processing. This can potentially enable energy-efficient hardware implementations of neuromorphic systems which emulate the functional units of the brain, namely, neurons and synapses. Recent demonstrations of ultra-fast photonic computing devices based on phase-change materials (PCMs) show promise of addressing limitations of electrically driven neuromorphic systems. However, scaling these standalone computing devices to a parallel in-memory computing primitive is a challenge. In this work, we utilize the optical properties of the PCM, Ge2Sb2Te5 (GST), to propose a Photonic Spiking Neural Network computing primitive, comprising of a non-volatile synaptic array integrated seamlessly with previously explored 'integrate-and-fire' neurons. The proposed design realizes an 'in-memory' computing platform that leverages the inherent parallelism of wavelength-division-multiplexing (WDM). We show that the proposed computing platform can be used to emulate a SNN inferencing engine for image classification tasks. The proposed design not only bridges the gap between isolated computing devices and parallel large-scale implementation, but also paves the way for ultra-fast computing and localized on-chip learning.

show abstract

Section: Introductionmentioning

confidence: 99%

Photonic In-Memory Computing Primitive for Spiking Neural Networks Using Phase-Change Materials

2019

View full text Add to dashboard Cite

show abstract

“…Png are with the Institute of High Performance Computing, Agency for Science, Technology and Research (A*STAR), 138632, Singapore. (e-mail: ongjr@ihpc.a-star.edu.sg; pngce@ihpc.a-star.edu.sg) networks are particularly suited for optical implementation as they mainly rely on computing matrix multiplications, which can be performed with high speed and throughput with optics [10], [11]. In fact, the inherent advantages of optical neural networks were studied in detail decades ago using bulk optics [12].…”

Section: Introductionmentioning

confidence: 99%

Photonic convolutional neural networks using integrated diffractive optics

Ong¹,

Ang²,

Ooi³

et al. 2020

Preprint

View full text Add to dashboard Cite

<div>With recent rapid advances in photonic integrated circuits, it has been demonstrated that programmable photonic chips can be used to implement artificial neural networks. Convolutional neural networks (CNN) are a class of deep learning methods that have been highly successful in applications such as image classification and speech processing. We present an architecture to implement a photonic CNN using the Fourier transform property of integrated star couplers. We show, in computer simulation, high accuracy image classification using the MNIST dataset. We also model component imperfections in photonic CNN and show that the performance degradation can be recovered in a programmable chip. Our proposed architecture provides a large reduction in physical footprint compared to current implementations as it utilizes the natural advantages of optics and hence offers a scalable pathway towards integrated photonic deep learning processors.</div>

show abstract

“…Typical microdisk lasers generate continuous wave (cw) emissions, which are guided by optical waveguides and distributed to multiple ends through Y-branching splitters, followed by the translations from cw beams to pulsed signals via external modulators [6][7][8]. Recently, due to the great demand for low energy consumption and high integration density, combining the laser and modulator to form a directly modulated light source has become a rapidly developing research direction in next-generation on-chip PCs [1,2].…”

Section: Introductionmentioning

confidence: 99%

Direct modulation of microcavity emission via local perturbation

Xiao

Shuai

et al. 2013

Phys. Rev. A

View full text Add to dashboard Cite

Next-generation integrated photonic circuits (PCs) require the combination of on-chip coherent light sources and modulators to increase the integration density and reduce the energy consumption. While significant advances have been accomplished in direct laser modulation and channeling the modulated emissions into waveguides, there is still an enormous challenge in distributing the signals to their designed ends. Here we show that the outputs from microcavities can be directly modulated by local perturbations. We demonstrate this concept in a waveguides connected quadruple cavity, where the generated signals can be optionally distributed to one or multiple designed waveguides by slightly modulating the refractive index on the order of 0.01. This concept should help the advances of high-density on-chip photonic circuits.

show abstract

On-chip CMOS-compatible optical signal processor

Cited by 79 publications

References 22 publications

Photonic In-Memory Computing Primitive for Spiking Neural Networks Using Phase-Change Materials

Photonic In-Memory Computing Primitive for Spiking Neural Networks Using Phase-Change Materials

Photonic convolutional neural networks using integrated diffractive optics

Direct modulation of microcavity emission via local perturbation

Contact Info

Product

Resources

About