Details
Originalsprache | Englisch |
---|---|
Aufsatznummer | 3020 |
Seitenumfang | 8 |
Fachzeitschrift | Nature Communications |
Jahrgang | 15 |
Ausgabenummer | 1 |
Frühes Online-Datum | 16 Apr. 2024 |
Publikationsstatus | Veröffentlicht - Dez. 2024 |
Extern publiziert | Ja |
Abstract
Recurrent neural networks (RNNs) can process contextual information such as time series signals and language. But their tracking of internal states is a limiting factor, motivating research on analog implementations in photonics. While photonic unidirectional feedforward neural networks (NNs) have demonstrated big leaps, bi-directional optical RNNs present a challenge: the need for a short-term memory that (i) programmable and coherently computes optical inputs, (ii) minimizes added noise, and (iii) allows scalability. Here, we experimentally demonstrate an optoacoustic recurrent operator (OREO) which meets (i, ii, iii). OREO contextualizes the information of an optical pulse sequence via acoustic waves. The acoustic waves link different optical pulses, capturing their information and using it to manipulate subsequent operations. OREO’s all-optical control on a pulse-by-pulse basis offers simple reconfigurability and is used to implement a recurrent drop-out and pattern recognition of 27 optical pulse patterns. Finally, we introduce OREO as bi-directional perceptron for new classes of optical NNs.
ASJC Scopus Sachgebiete
- Chemie (insg.)
- Allgemeine Chemie
- Biochemie, Genetik und Molekularbiologie (insg.)
- Allgemeine Biochemie, Genetik und Molekularbiologie
- Physik und Astronomie (insg.)
- Allgemeine Physik und Astronomie
Zitieren
- Standard
- Harvard
- Apa
- Vancouver
- BibTex
- RIS
in: Nature Communications, Jahrgang 15, Nr. 1, 3020, 12.2024.
Publikation: Beitrag in Fachzeitschrift › Artikel › Forschung › Peer-Review
}
TY - JOUR
T1 - An optoacoustic field-programmable perceptron for recurrent neural networks
AU - Becker, Steven
AU - Englund, Dirk
AU - Stiller, Birgit
N1 - Publisher Copyright: © The Author(s) 2024.
PY - 2024/12
Y1 - 2024/12
N2 - Recurrent neural networks (RNNs) can process contextual information such as time series signals and language. But their tracking of internal states is a limiting factor, motivating research on analog implementations in photonics. While photonic unidirectional feedforward neural networks (NNs) have demonstrated big leaps, bi-directional optical RNNs present a challenge: the need for a short-term memory that (i) programmable and coherently computes optical inputs, (ii) minimizes added noise, and (iii) allows scalability. Here, we experimentally demonstrate an optoacoustic recurrent operator (OREO) which meets (i, ii, iii). OREO contextualizes the information of an optical pulse sequence via acoustic waves. The acoustic waves link different optical pulses, capturing their information and using it to manipulate subsequent operations. OREO’s all-optical control on a pulse-by-pulse basis offers simple reconfigurability and is used to implement a recurrent drop-out and pattern recognition of 27 optical pulse patterns. Finally, we introduce OREO as bi-directional perceptron for new classes of optical NNs.
AB - Recurrent neural networks (RNNs) can process contextual information such as time series signals and language. But their tracking of internal states is a limiting factor, motivating research on analog implementations in photonics. While photonic unidirectional feedforward neural networks (NNs) have demonstrated big leaps, bi-directional optical RNNs present a challenge: the need for a short-term memory that (i) programmable and coherently computes optical inputs, (ii) minimizes added noise, and (iii) allows scalability. Here, we experimentally demonstrate an optoacoustic recurrent operator (OREO) which meets (i, ii, iii). OREO contextualizes the information of an optical pulse sequence via acoustic waves. The acoustic waves link different optical pulses, capturing their information and using it to manipulate subsequent operations. OREO’s all-optical control on a pulse-by-pulse basis offers simple reconfigurability and is used to implement a recurrent drop-out and pattern recognition of 27 optical pulse patterns. Finally, we introduce OREO as bi-directional perceptron for new classes of optical NNs.
UR - http://www.scopus.com/inward/record.url?scp=85190506800&partnerID=8YFLogxK
U2 - 10.1038/s41467-024-47053-6
DO - 10.1038/s41467-024-47053-6
M3 - Article
C2 - 38627394
AN - SCOPUS:85190506800
VL - 15
JO - Nature Communications
JF - Nature Communications
SN - 2041-1723
IS - 1
M1 - 3020
ER -