On-chip optical matrix-vector multiplier for parallel computation

Yang, Lin; Zhang, Lei; Ji, Ruiqiang

doi:10.1117/2.1201306.004932

Cited by 9 publications

(6 citation statements)

References 4 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Certain operations can be performed in free space or on a photonic chip with little to no power consumption, e.g. a lens can take a Fourier transform “for free” 11 , 12 . An optimizable and scalable set of optical configurations that preserves these advantages and serves as a framework for building optical CNNs would be of interest to computer vision, robotics, machine learning, and optics communities.…”

Section: Introductionmentioning

confidence: 99%

Hybrid optical-electronic convolutional neural networks with optimized diffractive optics for image classification

Chang

Sitzmann

Xiong

et al. 2018

Sci Rep

423

246

View full text Add to dashboard Cite

Convolutional neural networks (CNNs) excel in a wide variety of computer vision applications, but their high performance also comes at a high computational cost. Despite efforts to increase efficiency both algorithmically and with specialized hardware, it remains difficult to deploy CNNs in embedded systems due to tight power budgets. Here we explore a complementary strategy that incorporates a layer of optical computing prior to electronic computing, improving performance on image classification tasks while adding minimal electronic computational cost or processing time. We propose a design for an optical convolutional layer based on an optimized diffractive optical element and test our design in two simulations: a learned optical correlator and an optoelectronic two-layer CNN. We demonstrate in simulation and with an optical prototype that the classification accuracies of our optical systems rival those of the analogous electronic implementations, while providing substantial savings on computational cost.

show abstract

Section: Introductionmentioning

confidence: 99%

Hybrid optical-electronic convolutional neural networks with optimized diffractive optics for image classification

Chang

Sitzmann

Xiong

et al. 2018

Sci Rep

423

246

View full text Add to dashboard Cite

show abstract

“…1) Photonic matrix-vector multiplier: A microdisk-based photonic MVM relies on an array of microdisks, photodetectors (PDs) and a wavelength multiplexer to conduct parallel matrix-vector multiplications. Unlike a recent microring-based counterpart [25], we adopt microdisks in an MVM to reduce its area overhead and power consumption. 2) Photonic adder: We adopt a 16-bit ripple-carry adder consisting of 16 microdisk-based 1-bit EO (electro-optic) full adders [27], [28].…”

Section: A Computing Unit Construction and Modelingmentioning

confidence: 99%

“…A smaller microdisk array in a photonic MVM reduces its computing throughput, while a larger microdisk array degrades the MVM accuracy. Based on our photonic MVM simulations following the MVM design space exploration in [25], we conservatively adopted a 128 × 128 microdisk array in our photonic MVM to balance the tradeoff. Although a microring-based MVM [25] suggests each weight bank microring can store > 4 bits, we found large MVM accuracy degradations appear in our microdiskbased MVM when using the same configuration.…”

Section: A Computing Unit Construction and Modelingmentioning

confidence: 99%

“…Based on our photonic MVM simulations following the MVM design space exploration in [25], we conservatively adopted a 128 × 128 microdisk array in our photonic MVM to balance the tradeoff. Although a microring-based MVM [25] suggests each weight bank microring can store > 4 bits, we found large MVM accuracy degradations appear in our microdiskbased MVM when using the same configuration. Therefore, we conservatively use the transmissivity of each microdisk to represent only 2 bits of a 16-bit weight to maintain no MVM accuracy degradation.…”

Section: A Computing Unit Construction and Modelingmentioning

confidence: 99%

See 1 more Smart Citation

HolyLight: A Nanophotonic Accelerator for Deep Learning in Data Centers

Liu

et al. 2019

2019 Design, Automation &Amp; Test in Europe Conference &Amp; Exhibition (DATE)

View full text Add to dashboard Cite

Convolutional Neural Networks (CNNs) are widely adopted in object recognition, speech processing and machine translation, due to their extremely high inference accuracy. However, it is challenging to compute massive computationally expensive convolutions of deep CNNs on traditional CPUs and GPUs. Emerging Nanophotonic technology has been employed for on-chip data communication, because of its CMOS compatibility, high bandwidth and low power consumption. In this paper, we propose a nanophotonic accelerator, HolyLight, to boost the CNN inference throughput in datacenters. Instead of an all-photonic design, HolyLight performs convolutions by photonic integrated circuits, and process the other operations in CNNs by CMOS circuits for high inference accuracy. We first build HolyLight-M by microdisk-based matrix-vector multipliers. We find analog-todigital converters (ADCs) seriously limit its inference throughput per Watt. We further use microdisk-based adders and shifters to architect HolyLight-A without ADCs. Compared to the stateof-the-art ReRAM-based accelerator, HolyLight-A improves the CNN inference throughput per Watt by 13× with trivial accuracy degradation.

show abstract

“…The advantages of photonics have promoted extensive demonstrations of various high-performance optical functionalities, especially in photon computing [10][11][12][13][14][15]. The programmable Mach Zehnder interferometers realize optical interference units for optical neural networks [11,16], and MRRs are used for optical matrixvector multipliers to implement neuromorphic photonic networks [17][18][19]. In addition, diffractive optics [20][21][22], space-light modulators [23,24], semiconductor optical amplifiers [25], optical modulators [26,27], and other related optical devices are used to achieve powerful deep learning accelerators, which can perform machine learning tasks with ultra-low energy consumption [28].…”

Section: Introductionmentioning

confidence: 99%

Implementation of Pruned Backpropagation Neural Network Based on Photonic Integrated Circuits

2021

View full text Add to dashboard Cite

We demonstrate a pruned high-speed and energy-efficient optical backpropagation (BP) neural network. The micro-ring resonator (MRR) banks, as the core of the weight matrix operation, are used for large-scale weighted summation. We find that tuning a pruned MRR weight banks model gives an equivalent performance in training with the model of random initialization. Results show that the overall accuracy of the optical neural network on the MNIST dataset is 93.49% after pruning six-layer MRR weight banks on the condition of low insertion loss. This work is scalable to much more complex networks, such as convolutional neural networks and recurrent neural networks, and provides a potential guide for truly large-scale optical neural networks.

show abstract

On-chip optical matrix-vector multiplier for parallel computation

Cited by 9 publications

References 4 publications

Hybrid optical-electronic convolutional neural networks with optimized diffractive optics for image classification

Hybrid optical-electronic convolutional neural networks with optimized diffractive optics for image classification

HolyLight: A Nanophotonic Accelerator for Deep Learning in Data Centers

Implementation of Pruned Backpropagation Neural Network Based on Photonic Integrated Circuits

Contact Info

Product

Resources

About