ML approach classifies and generates structured light in turbulence · arXiv research

Researchers have developed a machine learning approach to classify and generate structured light beams that have propagated through turbulent media. The study utilizes numerical simulations to create speckle patterns and employs SimpleCNN and ResNet-18 classifiers to analyze intensity and autocorrelation inputs. To address the cost of obtaining additional propagated samples, a class-conditioned diffusion model was created for generative augmentation, featuring a spectrum-aware diffusion objective that combines pixel-domain loss with a Fourier-domain regularizer to maintain high-frequency speckle statistics. This hybrid objective has been shown to improve classification performance, particularly in low-data scenarios. AI

IMPACT This research could lead to improved methods for analyzing and simulating light propagation in complex environments, potentially impacting fields like optical communication and remote sensing.

RANK_REASON The cluster contains an academic paper detailing a new ML-based approach for a physics problem. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• Anjali Nair
• ResNet-18
• SimpleCNN

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →