Parameter-efficient neural computing with programmable photonic meshes
Abstract
Memory capacity and bandwidth are becoming major constraints in neural-network computing, motivating parameter-efficient representations that reduce model storage and data movement. Tensor-network methods already enable substantial compression of neural networks, but their structured operations are often poorly matched to conventional processors optimized for dense linear algebra. Such representations can be implemented in photonic hardware, although this typically requires optical architectures tailored to a particular factorization. Here we propose a different approach based on programmable photonic meshes. Compact electronic parameters determine the interferometer phases, while optical interference produces a dense, generally high-rank transformation. Since the parameterization remains electronic, its dimensionality and structure can be changed without modifying the optical circuit. Numerical experiments in end-to-end training demonstrate that the proposed photonic layers match or outperform strongly compressed tensor-train and brickwork layers, often with fewer trainable parameters. Experiments in parameter-efficient fine-tuning of a language model also show advantages over established digital adapters. The reduced trainable state lowers persistent memory requirements and offers the potential to reduce data movement and bandwidth demands.
Related
- Parameter-Efficient Multilinear Readout for Quantum Reservoir Computing
- Interferometric Readout of Momentum-Space Topology in a Programmable Dissipative Photonic Circuit
- Nonlinear pluggable optics: Digital signal processing-free Intensity Modulated Direct Detection links using analog photonic Next Generation Reservoir Computing
- Programmable Non-Hermitian Synchronization of Light on a Silicon Photonic Processor
- A programmable quantum photonic platform integrating coherent emitters
BibTeX
@misc{popov2026parameterefficient,
title = {{Parameter-efficient neural computing with programmable photonic meshes}},
author = {Artem Popov and Stanislav Straupe and Mikhail Saygin},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.26429},
eprint = {2610.26429},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.26429},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Parameter-efficient neural computing with programmable photonic meshes
AU - Artem Popov
AU - Stanislav Straupe
AU - Mikhail Saygin
PY - 2026
DA - 2026/10/19
PB - arXiv
DO - 10.48550/arxiv.2610.26429
UR - https://arxiv.org/abs/2610.26429
AB - Memory capacity and bandwidth are becoming major constraints in neural-network computing, motivating parameter-efficient representations that reduce model storage and data movement. Tensor-network methods already enable substantial compression of neural networks, but their structured operations are often poorly matched to conventional processors optimized for dense linear algebra. Such representations can be implemented in photonic hardware, although this typically requires optical architectures tailored to a particular factorization. Here we propose a different approach based on programmable photonic meshes. Compact electronic parameters determine the interferometer phases, while optical interference produces a dense, generally high-rank transformation. Since the parameterization remains electronic, its dimensionality and structure can be changed without modifying the optical circuit. Numerical experiments in end-to-end training demonstrate that the proposed photonic layers match or outperform strongly compressed tensor-train and brickwork layers, often with fewer trainable parameters. Experiments in parameter-efficient fine-tuning of a language model also show advantages over established digital adapters. The reduced trainable state lowers persistent memory requirements and offers the potential to reduce data movement and bandwidth demands.
ER -