Neuromorphic rate-coded representation of digital image data with a VCSEL-neuron

Matej Hejda; Joshua Robertson; Julian Bueno; Juan Alanis; Antonio Hurtado

Neuromorphic rate-coded representation of digital image data with a VCSEL-neuron

Matej Hejda, Joshua Robertson, Julian Bueno, Juan Alanis, Antonio Hurtado

Research output: Contribution to conference › Poster › peer-review

Abstract

Neuromorphic photonic approaches, focusing on implementation of brain-inspired hardware, have recently gained significant interest. Among those, vertical cavity surface emitting lasers (VCSELs) offer multiple desirable properties while exhibiting fast spiking responses. We utilize their functionality to experimentally realize temporal spike-based encoding of pixel intensity values from digital images. The image data is represented as continuous, rate-coded optical spike trains. Reconstruction from the spike rates shows very good agreement with the source images.

Original language	English
Publication status	Published - 4 Dec 2020
Event	European Semiconductor Laser Workshop - Virtual, Eindhoven, Netherlands Duration: 4 Dec 2020 → 5 Dec 2020 Conference number: 20

Conference

Conference	European Semiconductor Laser Workshop
Abbreviated title	ESLW
Country/Territory	Netherlands
City	Eindhoven
Period	4/12/20 → 5/12/20

Keywords

neuromorphic
rate-coded representation
digital image data
VCSEL-neuron

Energy-efficient and high-bandwidth neuromorphic nanophotonic chips for artificial intelligence systems (ChipAI) H2020-FETOPEN
Hurtado, A. & Strain, M.
European Commission - Horizon Europe + H2020
1/03/19 → 28/02/22
Project: Research
BRAIN-inspired networks of ultrafast LASER neurons
Hurtado, A.
Office of Naval Research ONR
10/11/17 → 9/11/21
Project: Research

Cite this

@conference{b5aa3e1dbf3646dcb3fea3c11b174593,

title = "Neuromorphic rate-coded representation of digital image data with a VCSEL-neuron",

abstract = "Neuromorphic photonic approaches, focusing on implementation of brain-inspired hardware, have recently gained significant interest. Among those, vertical cavity surface emitting lasers (VCSELs) offer multiple desirable properties while exhibiting fast spiking responses. We utilize their functionality to experimentally realize temporal spike-based encoding of pixel intensity values from digital images. The image data is represented as continuous, rate-coded optical spike trains. Reconstruction from the spike rates shows very good agreement with the source images.",

keywords = "neuromorphic, rate-coded representation, digital image data, VCSEL-neuron",

author = "Matej Hejda and Joshua Robertson and Julian Bueno and Juan Alanis and Antonio Hurtado",

year = "2020",

month = dec,

day = "4",

language = "English",

note = "European Semiconductor Laser Workshop, ESLW ; Conference date: 04-12-2020 Through 05-12-2020",

}

TY - CONF

T1 - Neuromorphic rate-coded representation of digital image data with a VCSEL-neuron

AU - Hejda, Matej

AU - Robertson, Joshua

AU - Bueno, Julian

AU - Alanis, Juan

AU - Hurtado, Antonio

N1 - Conference code: 20

PY - 2020/12/4

Y1 - 2020/12/4

N2 - Neuromorphic photonic approaches, focusing on implementation of brain-inspired hardware, have recently gained significant interest. Among those, vertical cavity surface emitting lasers (VCSELs) offer multiple desirable properties while exhibiting fast spiking responses. We utilize their functionality to experimentally realize temporal spike-based encoding of pixel intensity values from digital images. The image data is represented as continuous, rate-coded optical spike trains. Reconstruction from the spike rates shows very good agreement with the source images.

AB - Neuromorphic photonic approaches, focusing on implementation of brain-inspired hardware, have recently gained significant interest. Among those, vertical cavity surface emitting lasers (VCSELs) offer multiple desirable properties while exhibiting fast spiking responses. We utilize their functionality to experimentally realize temporal spike-based encoding of pixel intensity values from digital images. The image data is represented as continuous, rate-coded optical spike trains. Reconstruction from the spike rates shows very good agreement with the source images.

KW - neuromorphic

KW - rate-coded representation

KW - digital image data

KW - VCSEL-neuron

M3 - Poster

T2 - European Semiconductor Laser Workshop

Y2 - 4 December 2020 through 5 December 2020

ER -

Neuromorphic rate-coded representation of digital image data with a VCSEL-neuron

Abstract

Conference

Keywords

Fingerprint

Projects

Energy-efficient and high-bandwidth neuromorphic nanophotonic chips for artificial intelligence systems (ChipAI) H2020-FETOPEN

BRAIN-inspired networks of ultrafast LASER neurons

Cite this