English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Silicon-photonics-based optical neural network (ONN) is a promising hardwareplatform that could represent a paradigm shift in efficient AI with itsCMOS-compatibility, flexibility, ultra-low execution latency, and high energyefficiency. In-situ training on the online programmable photonic chips isappealing but still encounters challenging issues in on-chip implementability,scalability, and efficiency. In this work, we propose a closed-loop ONN on-chiplearning framework L2ight to enable scalable ONN mapping and efficient in-situlearning. L2ight adopts a three-stage learning flow that first calibrates thecomplicated photonic circuit states under challenging physical constraints,then performs photonic core mapping via combined analytical solving andzeroth-order optimization. A subspace learning procedure with multi-levelsparsity is integrated into L2ight to enable in-situ gradient evaluation andfast adaptation, unleashing the power of optics for real on-chip intelligence.Extensive experiments demonstrate our proposed L2ight outperforms prior ONNtraining protocols with 3-order-of-magnitude higher scalability and over 30Xbetter efficiency, when benchmarked on various models and learning tasks. Thissynergistic framework is the first scalable on-chip learning solution thatpushes this emerging field from intractable to scalable and further toefficient for next-generation self-learnable photonic neural chips. From aco-design perspective, L2ight also provides essential insights forhardware-restricted unitary subspace optimization and efficient sparsetraining. We open-source our framework athttps://github.com/JeremieMelo/L2ight.

L2ight: Enabling On-Chip Learning for Optical Neural Networks via Efficient in-situ Subspace Optimization