Scholarlink

This study investigates the integration of machine learning techniques (ML) to predict the mode effective indices in silicon waveguides. A comprehensive dataset was generated encompassing various parameters including waveguide dimensions, mode polarization, and effective refractive index. Through hyperparameter tuning process of different ML models, we demonstrated that ML techniques can reach high performance achieving an absolute percentage error of less than 5% across various designs while outperforming the numerical methods in term of calculation rapidity. These findings highlight the substantial potential of ML for optimizing silicon photonic components.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Optimizing Machine Learning Techniques for Precise Prediction of Optical Properties in Silicon Waveguides

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Mohamed Mammeri,
  • Babak Hashemi,
  • Teresa Crisci,
  • Maurizio Casalino,
  • Francesco Giuseppe Della Corte

摘要

This study investigates the integration of machine learning techniques (ML) to predict the mode effective indices in silicon waveguides. A comprehensive dataset was generated encompassing various parameters including waveguide dimensions, mode polarization, and effective refractive index. Through hyperparameter tuning process of different ML models, we demonstrated that ML techniques can reach high performance achieving an absolute percentage error of less than 5% across various designs while outperforming the numerical methods in term of calculation rapidity. These findings highlight the substantial potential of ML for optimizing silicon photonic components.