Early Classification of Time Series in Non-Stationary Cost Regimes

2026-01-31

AI Summary

This study addresses a critical yet overlooked challenge in time series early classification: the non-stationary evolution of decision costs during deployment—such as shifting trade-offs between misclassification and delay penalties—which often leads to a mismatch between training objectives and real-world performance. The work presents the first systematic investigation of this issue and proposes an online adaptation strategy that, while keeping the base classifier fixed, dynamically updates only the triggering model to respond to cost drift. Leveraging a decoupled early classification architecture, the approach integrates reinforcement learning and multi-armed bandit algorithms. Controlled experiments on synthetic data demonstrate that the proposed method significantly enhances robustness against non-stationary cost structures, with reinforcement learning exhibiting particularly stable and superior performance across diverse drift scenarios.

Technology Category

• Machine Learning: Online Learning & Bandits
• Search and Optimization: Learning to Search
• Reasoning under Uncertainty: Sequential Decision Making

Application Category

• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs

Abstract

Early Classification of Time Series (ECTS) addresses decision-making problems in which predictions must be made as early as possible while maintaining high accuracy. Most existing ECTS methods assume that the time-dependent decision costs governing the learning objective are known, fixed, and correctly specified. In practice, however, these costs are often uncertain and may change over time, leading to mismatches between training-time and deployment-time objectives. In this paper, we study ECTS under two practically relevant forms of cost non-stationarity: drift in the balance between misclassification and decision delay costs, and stochastic realizations of decision costs that deviate from the nominal training-time model. To address these challenges, we revisit representative ECTS approaches and adapt them to an online learning setting. Focusing on separable methods, we update only the triggering model during deployment, while keeping the classifier fixed. We propose several online adaptations and baselines, including bandit-based and RL-based approaches, and conduct controlled experiments on synthetic data to systematically evaluate robustness under cost non-stationarity. Our results demonstrate that online learning can effectively improve the robustness of ECTS methods to cost drift, with RL-based strategies exhibiting strong and stable performance across varying cost regimes.

Problem

• Early Classification of Time Series
• Non-Stationary Costs
• Cost Drift
• Decision Delay
• Misclassification Cost

Innovation

• Early Classification of Time Series
• Non-Stationary Costs
• Online Learning
• Reinforcement Learning
• Cost Drift