Hybrid Active-Online Learning Framework for Label-Efficient Concept Drift Adaptation in Optical Network Failure Detection

2026-06-29

AI Summary

This work addresses the challenges of model degradation due to concept drift and high labeling costs in optical network fault detection by proposing a hybrid framework that integrates active learning with online learning. The approach combines margin-based selective annotation, an online updating mechanism, and adaptive concept drift detection to achieve near-optimal accuracy and AUC scores while annotating only 3.4% of the most informative samples in the data stream. Experimental results demonstrate that the proposed method substantially reduces labeling overhead without compromising detection performance and incurs negligible additional inference latency, thereby enabling efficient and accurate real-time fault detection.

Abstract

We propose a hybrid active-online learning framework for label-efficient concept drift adaptation in optical network failure detection. Using margin-based selective labeling, our method achieves nearceiling accuracy and AUC scores while querying only 3.4% of streaming samples, with negligible latency overhead compared to static inference.

Problem

• concept drift
• label efficiency
• optical network failure detection
• active learning
• online learning

Innovation

• hybrid active-online learning
• label-efficient learning
• concept drift adaptation
• margin-based selective labeling
• optical network failure detection