Machine Learning for Optical Fiber Sensing: Generalization, Adaptation, and Real-World Deployment

Machine learning has shown strong potential for optical fiber sensing, yet many advances have been demonstrated primarily in controlled laboratory or testbed settings. Moving toward reliable field deployment requires models that can learn from limited supervision, generalize across different sensing environments and configurations, and adapt efficiently to new deployment scenarios and tasks. This tutorial will examine how modern machine learning techniques can address these challenges, with an emphasis on generalization, adaptation, and practical considerations in real-world deployment. Using distributed fiber-optic sensing on deployed telecommunication fiber networks as primary examples, we will discuss inductive biases, weakly supervised and self-supervised learning, generative AI, foundation-model adaptation, and practical use cases. The tutorial will connect recent advances in machine learning with lessons from field deployments and highlight emerging opportunities for building more generalizable, adaptable, and efficient sensing-AI systems.

Nov 30, 2026