Work

20 GHz In-situ Optical Backpropagation for Training AWGR-based Hyperdimensional Photonic Tensor Cores

Year: 2026

Type: article

Abstract: We demonstrate in-situ optical backpropagation and train a hyperdimensional AWGR-based photonic tensor core, performing both forward and backward MVM operations at 20 GHz and presenting inference accuracies of 93.5% and 81.6% for the Wine and Moons datasets, respectively.

Source: Zenodo (CERN European Organization for Nuclear Research)

DOI: 10.5281/zenodo.24103607

Authors: Asimopoulos, Odysseas, Kovaios, S., Moralis - Pegios, M., Avramelou, L., Passalis, N., Tefas, A., Vyrskokinos, K., Pleros, N.

Institution: Aristotle University of Thessaloniki

Language: English

Topic: Neural Networks and Reservoir Computing

Subfield: Artificial Intelligence

Field: Computer Science

Domain: Physical Sciences

Open Access status: green

OpenAlex ID: W7225733994

FWCI:

Cites:

Cited by:

Related to: