2022 IEEE Photonics Conference (IPC) 2022
DOI: 10.1109/ipc53466.2022.9975526
|View full text |Cite
|
Sign up to set email alerts
|

Reducing Latency in Sensing for Optical Convolutional Neural Network

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
5

Citation Types

0
7
0

Year Published

2023
2023
2023
2023

Publication Types

Select...
7

Relationship

4
3

Authors

Journals

citations
Cited by 9 publications
(7 citation statements)
references
References 5 publications
0
7
0
Order By: Relevance
“…After two decades development, kernels generated by state-in-the-art DMD outperforms the older hologram mask with a higher enough throughput to simulate completely neural planes, makes optical convolutional efficiency promising. [87][88][89][90][91][92][93][94][95][96][97][98][99][100][101] In this study, we experimentally generate multiplexed OAM beams and introduce optical filtering method which relys on spatial Fourier transform of images in the frequency domain as optical convolutional neural network to train and identify multiplexed OAM beams under simulated atmospheric turbulence conditions, we show the system currently capable of classifying 12 classes at test accuracy of 95% (under weak turbulence) and 87% (under strong turbulence).…”
Section: Introductionmentioning
confidence: 97%
“…After two decades development, kernels generated by state-in-the-art DMD outperforms the older hologram mask with a higher enough throughput to simulate completely neural planes, makes optical convolutional efficiency promising. [87][88][89][90][91][92][93][94][95][96][97][98][99][100][101] In this study, we experimentally generate multiplexed OAM beams and introduce optical filtering method which relys on spatial Fourier transform of images in the frequency domain as optical convolutional neural network to train and identify multiplexed OAM beams under simulated atmospheric turbulence conditions, we show the system currently capable of classifying 12 classes at test accuracy of 95% (under weak turbulence) and 87% (under strong turbulence).…”
Section: Introductionmentioning
confidence: 97%
“…Artificial neurons based on nanophotonic technologies can potentially provide the platform that can fulfill the challenging future technological needs. [1][2][3][4][5][6][7][8][9][10][11][12][13][14][15][16][17][18][19][20] Integrated photonic technology provides a solution to the limitations of current digital electronic counterparts like efficient fundamental computational operations such as weighted sum or addition, vector matrix multiplications, or convolutions technologically enabled by attojoule efficient electro-optic (EO) modulators, phase shifters, and combiners. Furthermore, high parallelism and bandwidth is provided by exploiting wavelength-, polarization-and/or mode-division multiplexing.…”
Section: Introductionmentioning
confidence: 99%
“…[35][36][37][38][39][40] Laws of quantum mechanics limit the Speed of classical electronic semiconductor hardware-based complicated calculations. [41][42][43][44][45][46][47][48][49] At the same time, in optics, the Optical Fourier Transform (FT) and dot-product multiplication performed passively by a single lens or metalens have reduced computational complexity compared to logarithmic scaling in electronic processing units, [50][51][52][53][54][55][56] both free-space [57][58][59][60][61][62][63][64][65] and integrated photonic integrated circuits versions. [66][67][68][69][70][71][72][73][74][75][76][77][78][79][80] The only drawback of Spatial Light Modulators (SLMs) based optical computing systems would be the refresh rate of the device itself, but with recent resea...…”
Section: Introductionmentioning
confidence: 99%