Programmable Synapses and Dendritic Circuits for Superconducting Optoelectronic Neuromorphic Computing

Superconducting optoelectronic hardware is promising for large-scale neuromorphic computing. In this work, analog circuits combining Josephson junctions and superconducting single-photon detectors are fabricated and shown to exhibit a variety of neuromorphic functions. First, single-photon sensitive...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:2024 International Conference on Neuromorphic Systems (ICONS) S. 277 - 281
Hauptverfasser: Primavera, Bryce A., Khan, Saeed, Adler, Samuel R., Shainline, Jeffrey M.
Format: Tagungsbericht
Sprache:Englisch
Veröffentlicht: IEEE 30.07.2024
Schlagworte:
Computational efficiency
Detectors
Hardware
Josephson junctions
Neuromorphic engineering
Optics
optoelectronic
Performance evaluation
Programming
spiking
superconducting
Synapses
System-on-chip
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:Superconducting optoelectronic hardware is promising for large-scale neuromorphic computing. In this work, analog circuits combining Josephson junctions and superconducting single-photon detectors are fabricated and shown to exhibit a variety of neuromorphic functions. First, single-photon sensitive synapses are demonstrated with local programmable memory cells. The high programming speed and low programming energies of these devices make them well-suited for future implementations of online and on-chip learning. Second, dendritic circuits are presented as a first demonstration of fan-in in superconducting optoelectronic hardware. These circuits perform biologically inspired operations including spike coincidence and sequence detection that are central to both the computational abilities of dendritic arbors and plasticity operations. Together, these advances underscore the suitability of this hardware for highly parallelized on-chip learning and sophisticated neural processing.
DOI:10.1109/ICONS62911.2024.00048