Scholarlink

<p>This work presents a phase-accumulation-based machine learning readout for dynamically perturbed graphene terahertz resonators. Instead of relying on steady-state or peak-only observables, the proposed approach integrates the transient phase response to construct physically meaningful features that capture both perturbation magnitude and temporal behavior. A physics-driven synthetic dataset is generated using a reduced-order dynamic model, incorporating parameter variability and noise. Multiple feature configurations are evaluated for classification and regression tasks. The results show that accumulated-phase descriptors alone are insufficient, but significantly improve performance when combined with instantaneous features, leading to enhanced robustness and predictive accuracy, particularly under low signal-to-noise conditions. The proposed framework provides a physics-informed pathway for integrating machine learning into graphene-based terahertz sensing systems.</p>

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

Phase accumulation for machine learning readout in graphene terahertz resonators

  • 期刊论文

  • 发表日期：

    2026年7月6日

查看全文
  • Álvaro Garay

摘要

This work presents a phase-accumulation-based machine learning readout for dynamically perturbed graphene terahertz resonators. Instead of relying on steady-state or peak-only observables, the proposed approach integrates the transient phase response to construct physically meaningful features that capture both perturbation magnitude and temporal behavior. A physics-driven synthetic dataset is generated using a reduced-order dynamic model, incorporating parameter variability and noise. Multiple feature configurations are evaluated for classification and regression tasks. The results show that accumulated-phase descriptors alone are insufficient, but significantly improve performance when combined with instantaneous features, leading to enhanced robustness and predictive accuracy, particularly under low signal-to-noise conditions. The proposed framework provides a physics-informed pathway for integrating machine learning into graphene-based terahertz sensing systems.