PreprintCC BY 4.0arXiv
Neural Networks with Exciton-Polariton Waveguides
Abstract
We propose and numerically study a scheme for neural networks using exciton-polariton waveguide meshes. These enable arbitrary programmable interlayer couplings suited to a restricted Boltzmann machine architecture where polariton phases act as stochastic network variables. We show that such a network can be trained using unsupervised Hebbian and anti-Hebbian updates. This allows the implementation of an associative memory, where the network recovers learned patterns from imperfect inputs.
Related
- Exciton-Tunable Phase Control and Superstrong Coupling Through Multimode Polariton Engineering in Planar Waveguides
- Spin Simulators based on Exciton-Polariton Waveguide Networks
- Observation of a supersolid phase in a spin-orbit coupled exciton-polariton Bose-Einstein condensate
- Nano-wrinkle waveguides for programmable polariton channeling.
- Exciton-polariton condensates in epsilon-near-zero cavities
BibTeX
@misc{xu2026neural,
title = {{Neural Networks with Exciton-Polariton Waveguides}},
author = {Huawen Xu and Ruiqi Bao and Lingyu Tian and Yusong Gan and Junfeng Gao and Michal Matuszewski and Qihua Xiong and Timothy Liew},
howpublished = {arXiv},
year = {2026},
month = nov,
doi = {10.48550/arxiv.2611.05846},
eprint = {2611.05846},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2611.05846},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Neural Networks with Exciton-Polariton Waveguides
AU - Huawen Xu
AU - Ruiqi Bao
AU - Lingyu Tian
AU - Yusong Gan
AU - Junfeng Gao
AU - Michal Matuszewski
AU - Qihua Xiong
AU - Timothy Liew
PY - 2026
DA - 2026/11/04
PB - arXiv
DO - 10.48550/arxiv.2611.05846
UR - https://arxiv.org/abs/2611.05846
AB - We propose and numerically study a scheme for neural networks using exciton-polariton waveguide meshes. These enable arbitrary programmable interlayer couplings suited to a restricted Boltzmann machine architecture where polariton phases act as stochastic network variables. We show that such a network can be trained using unsupervised Hebbian and anti-Hebbian updates. This allows the implementation of an associative memory, where the network recovers learned patterns from imperfect inputs.
ER -