DT-GOL: Dual-Track Geometric Online Learning in Nonstationary Environment with Label Delay
2026-06-22 • Machine Learning
Machine Learning
AI summaryⓘ
The authors address the problem of learning from data that changes over time but suffers from delayed labeling, which slows down adaptation. They propose a method called DT-GOL that uses the shape and structure of data to predict changes before labels arrive, treating the problem like semi-supervised learning. Instead of making hard guesses, their approach uses a soft labeling system that accounts for uncertainty to avoid mistakes. They also separate the learning into two parts: one that learns slowly from true labels and another that adapts quickly using the geometric clues. Their method works better than existing ones, especially when the data changes in complex ways.
online learningconcept driftlabel latencysemi-supervised learningpseudo-labelinggeometric surrogatestability-plasticity dilemmadual-track architecturetopological evolutionconfirmation bias
Authors
Yulin Wang, Yi He, Dianlong You, Di Wu
Abstract
Online learning is crucial for handling complex data streams in big data applications. Recent research has begun to focus on dynamic scenarios, i.e., non-stationary environments. However, a crucial yet often overlooked aspect is label latency, where new data may not receive labels in time due to the slow and expensive labeling process, thus hindering rapid adaptation to dynamic environments. To resolve this impasse, we propose Dual-Track Geometry Online Learning (DT-GOL), a novel framework that shifts from temporal compensation to spatial reasoning to bridge the supervised latency gap. By modeling the delay challenge as a semi-supervised task, we leverage real-time topological evolution of features as a reliable geometric surrogate for unobservable conceptual changes to achieve proactive supervised adaptation within the delay window. Unlike rigid self-training, we introduce a dynamic evidence calibration mechanism that distills geometric information into soft labels that perceive uncertainty, effectively mitigating the confirmation bias inherent in hard pseudo-labels. Furthermore, to resolve the stability-plasticity dilemma, we design a decoupled dual-track architecture in which a master learner serves as a stable anchor, updated strictly from delayed ground truth, while a transient branch leverages soft geometric knowledge for low-risk forward adaptation. Extensive experiments on real and synthetic datasets demonstrate that DT-GOL significantly outperforms existing state-of-the-art baseline methods, especially in scenarios with concept drift.