Photonic tensor core for machine learning: a review

Peserico, Nicola; Ma, Xiaoxuan; Shastri, Bahvin; Sorger, Volker J.

doi:10.1117/12.2633916

Cited by 6 publications

(4 citation statements)

References 40 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Such an enhanced FSR device holds potential applications in chemical sensing, all-optical wavelength division multiplexed networks, tunable filters or switches, biomolecule sensing, and environmental monitoring. 18 The proposed structure can also be significantly applicable in photonic neural networks (PNN) where micro-ring resonators function as weights 19 and can exhibit nonlinear activation properties. 20 Thus, by varying various device parameters, the weights of the PNN can be actively changed, hence by enhancing the performance of the PNN.…”

Section: Introductionmentioning

confidence: 99%

Integrated optical ring resonator based comb filter in silicon-on-insulator with through-to-add port feedback path showing larger free spectral range

Sah,

Kumar,

Pura Rangegowda

et al. 2024

J. Nanophoton.

View full text Add to dashboard Cite

A nested ring resonator structure has been devised on a silicon-on-insulator platform, incorporating a feedback path through the addition of a port to the device. This design selectively enhances resonant wavelengths while effectively suppressing numerous consecutive resonant wavelengths by ∼30 dB. Consequently, it significantly expands the free spectral range (FSR) by over 85 nm, exceeding 20 times the FSR achieved without the feedback path. The resonant peak wavelength can be tuned by ∼40 nm∕μm change in the feedback path, offering a broader range for refractive index sensing. The impact of the additional phase control parameter length in the feedback path was investigated. Theoretical analysis was followed by finite differential time domain simulations to validate the results. Various device parameters were systematically adjusted to observe a substantial increase in the FSR and improved controllability over peak transmission.

show abstract

Section: Introductionmentioning

confidence: 99%

Integrated optical ring resonator based comb filter in silicon-on-insulator with through-to-add port feedback path showing larger free spectral range

Sah,

Kumar,

Pura Rangegowda

et al. 2024

J. Nanophoton.

View full text Add to dashboard Cite

show abstract

“…Nonetheless, advancements in technology have highlighted several shortcomings of DMDs in both research and commercial applications, including issues in terms of size and weight, flexibility, and efficiency. Compared to other proposed on-chip integrated photonic hardware accelerators and in-memory computational photonic tensor cores, all-optical convolution solutions offer an alternative approach with a higher data processing rate and lower power consumption [ 5 , 6 , 7 , 8 , 9 ]. A 4f-type optical system has been utilized to introduce a method for optical matrix multiplication.…”

Section: Introductionmentioning

confidence: 99%

Enhancing Focusing and Defocusing Capabilities with a Dynamically Reconfigurable Metalens Utilizing Sb2Se3 Phase-Change Material

Shen

Peserico

et al. 2023

Nanomaterials

Self Cite

View full text Add to dashboard Cite

Metalenses are emerging as an alternative to digital micromirror devices (DMDs), with the advantages of compactness and flexibility. The exploration of metalenses has ignited enthusiasm among optical engineers, positioning them as the forthcoming frontier in technology. In this paper, we advocate for the implementation of the phase-change material, Sb2Se3, capable of providing swift, reversible, non-volatile focusing and defocusing within the 1550 nm telecom spectrum. The lens, equipped with a robust ITO microheater, offers unparalleled functionality and constitutes a significant step toward dynamic metalenses that can be integrated with beamforming applications. After a meticulously conducted microfabrication process, we showcase a device capable of rapid tuning (0.1 MHz level) for metalens focusing and defocusing at C band communication, achieved by alternating the PCM state between the amorphous and crystalline states. The findings from the experiment show that the device has a high contrast ratio for switching of 28.7 dB.

show abstract

“…Here, we develop and demonstrate a highly integrated Silicon Photonic Tensor Core [87][88][89][90][91][92][93][94][95][96]. By using photonic wire bonding [97][98][99][100], three tunable lasers are integrated into the circuit.…”

Section: Introductionmentioning

confidence: 99%

Photonic tensor core machine learning accelerator

Nouri²,

Pererico

et al. 2023

AI and Optical Data Sciences IV

View full text Add to dashboard Cite

Photonic Tensor Cores (PTCs) have been raised as one of the major candidates to accelerate Neural Network hardware, taking advantage of the high bandwidth, low latency, and energy efficiency that propagating light has over the electrical counterpart. However, actual solutions still rely on external bulk components, such as lasers, modulators, and photodetectors. In this work, we show the first fully integrated hybrid Silicon Photonics PTC for computing Matrix-Vector Multiplication (MVM), one of the main steps for any Neural Network layer. The PTC is formed by a WDM 3-fold InP laser array connected to an active Silicon Photonic PIC through Photonic Wire Bondings. The CW lasers are modulated by high-speed Mach-Zehnder modulators to generate the input vector. The mix of all the signals is sent into a 3x3 matrix, formed by highspeed add-drop coupled microring resonators, whose tuning signals are the weights of the matrix. Outputs are collected by a bank of high-speed integrated photodetectors. The whole photonics IC has a footprint of 4.1×1.7 mm 2 , including lasers, allowing to have just electrical I/O. The full integration of input modulators, weights, and photodetectors can allow the PIC to work at over 20 GHz bandwidth with extremely low latency. This integration is a key step toward the actual deployment of photonics as an NN-accelerator for future AI systems.

show abstract

Photonic tensor core for machine learning: a review

Cited by 6 publications

References 40 publications

Integrated optical ring resonator based comb filter in silicon-on-insulator with through-to-add port feedback path showing larger free spectral range

Integrated optical ring resonator based comb filter in silicon-on-insulator with through-to-add port feedback path showing larger free spectral range

Enhancing Focusing and Defocusing Capabilities with a Dynamically Reconfigurable Metalens Utilizing Sb2Se3 Phase-Change Material

Photonic tensor core machine learning accelerator

Contact Info

Product

Resources

About