New iPINN framework enhances phase retrieval in nonlinear spectroscopy

Researchers have developed an inverse physics-informed neural network (iPINN) designed to tackle the complex phase retrieval problem in broadband coherent anti-Stokes Raman spectroscopy (BCARS). This novel framework reconstructs the resonant susceptibility by predicting Lorentzian peak parameters from raw BCARS spectra, utilizing a differentiable analytical forward model. The iPINN incorporates a transformer encoder to identify spectral features and a multi-view consistency loss to ensure invariance across varying non-resonant background patterns, strengths, and noise levels. This approach demonstrates superior accuracy and robustness compared to direct spectral regression methods, achieving significantly lower error rates on benchmark datasets and maintaining accuracy across different measurement conditions. AI

影响 This research introduces a novel neural network architecture for complex spectroscopy problems, potentially improving data analysis in scientific research.

排序理由 The cluster contains an academic paper detailing a new method for a specific scientific problem. [lever_c_demoted from research: ic=1 ai=1.0]

在 arXiv cs.LG 阅读 →

• arXiv
• Broadband CARS spectral phase retrieval using a time-domain Kramers-Kronig transform
• Ipinnu
• Lorentzianthus
• Transformer encoder

• 论文
• 模型发布

AI 生成摘要 · Google Gemini · 来自 1 个来源。 我们如何撰写摘要 →