EUVIMEDEUVIMED entdecken
Crossref

Control-free and efficient integrated photonic neural networks via hardware-aware training and pruning

Tengji Xu, Weipeng Zhang, Jiawei Zhang, Zeyu Luo, Qiarong Xiao, Benshan Wang, Mingcheng Luo, Xingyuan Xu, Bhavin J. Shastri, Paul R. Prucnal, Chaoran Huang

Optica · 20.08.2024 · 11 (8) · 1039

DOI: 10.1364/optica.523225 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Optica, online erschienen am 23.07.2024. Verfasst von Tengji Xu und 10 weiteren Autor:innen. Die Arbeit wurde bisher 47-mal zitiert (Crossref) und nennt 61 Literaturangaben. Gefördert von: Innovation and Technology Fund, Shun Hing Institute of Advanced Engineering, Chinese University of Hong Kong.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Integrated photonic neural networks (PNNs) are at the forefront of AI computing, leveraging light’s unique properties, such as large bandwidth, low latency, and potentially low power consumption. Nevertheless, the integrated optical components are inherently sensitive to external disturbances, thermal interference, and various device imperfections, which detrimentally affect computing accuracy and reliability. Conventional solutions use complicated control methods to stabilize optical devices and chip, which result in high hardware complexity and are impractical for large-scale PNNs. To address this, we propose a training approach to enable control-free, accurate, and energy-efficient photonic computing without adding hardware complexity. The core idea is to train the parameters of a physical neural network towards its noise-robust and energy-efficient region. Our method is validated on different integrated PNN architectures and is applicable to solve various device imperfections in thermally tuned PNNs and PNNs based on phase change materials. A notable 4-bit improvement is achieved in micro-ring resonator-based PNNs without needing complex device control or power-hungry temperature stabilization circuits. Additionally, our approach reduces the energy consumption by tenfold. This advancement represents a significant step towards the practical, energy-efficient, and noise-resilient implementation of large-scale integrated PNNs.

Förderung

  • Innovation and Technology Fund
  • Shun Hing Institute of Advanced Engineering
  • Chinese University of Hong Kong
  • University Grants Committee

Zitieren

APA

Xu, T., Zhang, W., Zhang, J., Luo, Z., Xiao, Q., Wang, B., Luo, M., Xu, X., Shastri, B. J., Prucnal, P. R., & Huang, C. (2024). Control-free and efficient integrated photonic neural networks via hardware-aware training and pruning. Optica, 11(8), 1039. https://doi.org/10.1364/optica.523225

Vancouver

Xu T, Zhang W, Zhang J, Luo Z, Xiao Q, Wang B, et al. Control-free and efficient integrated photonic neural networks via hardware-aware training and pruning. Optica. 2024;11(8):1039. doi:10.1364/optica.523225

BibTeX
@article{Xu2024,
  author = {Xu, Tengji and Zhang, Weipeng and Zhang, Jiawei and Luo, Zeyu and Xiao, Qiarong and Wang, Benshan and Luo, Mingcheng and Xu, Xingyuan and Shastri, Bhavin J. and Prucnal, Paul R. and Huang, Chaoran},
  title = {Control-free and efficient integrated photonic neural networks via hardware-aware training and pruning},
  journal = {Optica},
  year = {2024},
  volume = {11},
  number = {8},
  pages = {1039},
  doi = {10.1364/optica.523225}
}
Kostenlos starten →