TUTCRIS - Tampereen teknillinen yliopisto

TUTCRIS

Neuromorphic Photonics with Coherent Linear Neurons Using Dual-IQ Modulation Cells

Tutkimustuotosvertaisarvioitu

Standard

Neuromorphic Photonics with Coherent Linear Neurons Using Dual-IQ Modulation Cells. / Mourgias-Alexandris, George; Totovic, Angelina; Tsakyridis, Apostolos; Passalis, Nikolaos; Vyrsokinos, Konstantinos; Tefas, Anastasios; Pleros, Nikos.

julkaisussa: Journal of Lightwave Technology, Vuosikerta 38, Nro 4, 2020, s. 811-819.

Tutkimustuotosvertaisarvioitu

Harvard

Mourgias-Alexandris, G, Totovic, A, Tsakyridis, A, Passalis, N, Vyrsokinos, K, Tefas, A & Pleros, N 2020, 'Neuromorphic Photonics with Coherent Linear Neurons Using Dual-IQ Modulation Cells', Journal of Lightwave Technology, Vuosikerta. 38, Nro 4, Sivut 811-819. https://doi.org/10.1109/JLT.2019.2949133

APA

Mourgias-Alexandris, G., Totovic, A., Tsakyridis, A., Passalis, N., Vyrsokinos, K., Tefas, A., & Pleros, N. (2020). Neuromorphic Photonics with Coherent Linear Neurons Using Dual-IQ Modulation Cells. Journal of Lightwave Technology, 38(4), 811-819. https://doi.org/10.1109/JLT.2019.2949133

Vancouver

Mourgias-Alexandris G, Totovic A, Tsakyridis A, Passalis N, Vyrsokinos K, Tefas A et al. Neuromorphic Photonics with Coherent Linear Neurons Using Dual-IQ Modulation Cells. Journal of Lightwave Technology. 2020;38(4):811-819. https://doi.org/10.1109/JLT.2019.2949133

Author

Mourgias-Alexandris, George ; Totovic, Angelina ; Tsakyridis, Apostolos ; Passalis, Nikolaos ; Vyrsokinos, Konstantinos ; Tefas, Anastasios ; Pleros, Nikos. / Neuromorphic Photonics with Coherent Linear Neurons Using Dual-IQ Modulation Cells. Julkaisussa: Journal of Lightwave Technology. 2020 ; Vuosikerta 38, Nro 4. Sivut 811-819.

Bibtex - Lataa

@article{62389b869ca24bdd8eabd1ef94ce0f32,
title = "Neuromorphic Photonics with Coherent Linear Neurons Using Dual-IQ Modulation Cells",
abstract = "Neuromorphic photonics aims to transfer the high-bandwidth and low-energy credentials of optics into neuromorphic computing architectures. In this effort, photonic neurons are trying to combine the optical interconnect segments with optics that can realize all critical constituent neuromorphic functions, including the linear neuron stage and the activation function. However, aligning this new platform with well-established neural network training models in order to allow for the synergy of the photonic hardware with the best-in-class training algorithms, the following requirements should apply: i) the linear photonic neuron has to be able to handle both positive and negative weight values, ii) the activation function has to closely follow the widely used mathematical activation functions that have already shown an enormous performance in demonstrated neural networks so far. Herein, we demonstrate a coherent linear neuron architecture that relies on a dual-IQ modulation cell as its basic neuron element, introducing distinct optical elements for weight amplitude and weight sign representation and exploiting binary optical carrier phase-encoding for positive/negative number representation. We present experimental results of a typical IQ modulator performing as an elementary two-input linear neuron cell and successfully implementing all-optical linear algebraic operations with 104-ps long optical pulses. We also provide the theoretical proof and formulation of how to extend a dual-IQ modulation cell into a complete N-input coherent linear neuron stage that requires only a single-wavelength optical input and avoids the resource-consuming Wavelength Division Multiplexing (WDM) weighting schemes. An 8-input coherent linear neuron is then combined with an experimentally validated optical sigmoid activation function into a physical layer simulation environment, with respective training and physical layer simulation results for the MNIST dataset revealing an average accuracy of 97.24{\%} and 94.37{\%}, respectively.",
keywords = "All-optical signal processing, neural networks, neuromorphic computing, neuromorphic photonics, optical neural network accelerators",
author = "George Mourgias-Alexandris and Angelina Totovic and Apostolos Tsakyridis and Nikolaos Passalis and Konstantinos Vyrsokinos and Anastasios Tefas and Nikos Pleros",
note = "EXT={"}Tefas, Anastasios{"}",
year = "2020",
doi = "10.1109/JLT.2019.2949133",
language = "English",
volume = "38",
pages = "811--819",
journal = "Journal of Lightwave Technology",
issn = "0733-8724",
publisher = "Optical Society of America",
number = "4",

}

RIS (suitable for import to EndNote) - Lataa

TY - JOUR

T1 - Neuromorphic Photonics with Coherent Linear Neurons Using Dual-IQ Modulation Cells

AU - Mourgias-Alexandris, George

AU - Totovic, Angelina

AU - Tsakyridis, Apostolos

AU - Passalis, Nikolaos

AU - Vyrsokinos, Konstantinos

AU - Tefas, Anastasios

AU - Pleros, Nikos

N1 - EXT="Tefas, Anastasios"

PY - 2020

Y1 - 2020

N2 - Neuromorphic photonics aims to transfer the high-bandwidth and low-energy credentials of optics into neuromorphic computing architectures. In this effort, photonic neurons are trying to combine the optical interconnect segments with optics that can realize all critical constituent neuromorphic functions, including the linear neuron stage and the activation function. However, aligning this new platform with well-established neural network training models in order to allow for the synergy of the photonic hardware with the best-in-class training algorithms, the following requirements should apply: i) the linear photonic neuron has to be able to handle both positive and negative weight values, ii) the activation function has to closely follow the widely used mathematical activation functions that have already shown an enormous performance in demonstrated neural networks so far. Herein, we demonstrate a coherent linear neuron architecture that relies on a dual-IQ modulation cell as its basic neuron element, introducing distinct optical elements for weight amplitude and weight sign representation and exploiting binary optical carrier phase-encoding for positive/negative number representation. We present experimental results of a typical IQ modulator performing as an elementary two-input linear neuron cell and successfully implementing all-optical linear algebraic operations with 104-ps long optical pulses. We also provide the theoretical proof and formulation of how to extend a dual-IQ modulation cell into a complete N-input coherent linear neuron stage that requires only a single-wavelength optical input and avoids the resource-consuming Wavelength Division Multiplexing (WDM) weighting schemes. An 8-input coherent linear neuron is then combined with an experimentally validated optical sigmoid activation function into a physical layer simulation environment, with respective training and physical layer simulation results for the MNIST dataset revealing an average accuracy of 97.24% and 94.37%, respectively.

AB - Neuromorphic photonics aims to transfer the high-bandwidth and low-energy credentials of optics into neuromorphic computing architectures. In this effort, photonic neurons are trying to combine the optical interconnect segments with optics that can realize all critical constituent neuromorphic functions, including the linear neuron stage and the activation function. However, aligning this new platform with well-established neural network training models in order to allow for the synergy of the photonic hardware with the best-in-class training algorithms, the following requirements should apply: i) the linear photonic neuron has to be able to handle both positive and negative weight values, ii) the activation function has to closely follow the widely used mathematical activation functions that have already shown an enormous performance in demonstrated neural networks so far. Herein, we demonstrate a coherent linear neuron architecture that relies on a dual-IQ modulation cell as its basic neuron element, introducing distinct optical elements for weight amplitude and weight sign representation and exploiting binary optical carrier phase-encoding for positive/negative number representation. We present experimental results of a typical IQ modulator performing as an elementary two-input linear neuron cell and successfully implementing all-optical linear algebraic operations with 104-ps long optical pulses. We also provide the theoretical proof and formulation of how to extend a dual-IQ modulation cell into a complete N-input coherent linear neuron stage that requires only a single-wavelength optical input and avoids the resource-consuming Wavelength Division Multiplexing (WDM) weighting schemes. An 8-input coherent linear neuron is then combined with an experimentally validated optical sigmoid activation function into a physical layer simulation environment, with respective training and physical layer simulation results for the MNIST dataset revealing an average accuracy of 97.24% and 94.37%, respectively.

KW - All-optical signal processing

KW - neural networks

KW - neuromorphic computing

KW - neuromorphic photonics

KW - optical neural network accelerators

U2 - 10.1109/JLT.2019.2949133

DO - 10.1109/JLT.2019.2949133

M3 - Article

VL - 38

SP - 811

EP - 819

JO - Journal of Lightwave Technology

JF - Journal of Lightwave Technology

SN - 0733-8724

IS - 4

ER -