Deep Reinforcement Learning based Triggering Function for Early Classifiers of Time Series

2025-02-10

AI Summary

This paper addresses the suboptimality of hand-crafted trigger functions in Early Classification of Time Series (ECTS). It is the first to systematically formulate ECTS as a reinforcement learning (RL) task, enabling end-to-end learning of dynamic early-termination policies. We propose a learnable and scalable state representation mechanism that jointly encodes temporal features and online observations to construct a compact, decision-oriented state space. An RL-based trigger—Alert—is designed using PPO or DQN. Evaluated on multiple benchmark datasets, our method achieves classification 42% earlier on average while improving accuracy by 3.7–9.2% over state-of-the-art methods, significantly outperforming heuristic triggers. Key contributions are: (1) a paradigm shift from heuristic design to RL-based ECTS formulation; (2) a learnable, engineered state-space design tailored for sequential decision-making; and (3) the first end-to-end optimized deep RL framework for early time-series classification.

Technology Category

• Machine Learning: Time-Series/Data Streams
• Search and Optimization: Learning to Search
• Intelligent Robots: State Estimation

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Economics, Online Markets and Human Computation: Incentives in network design for Web infrastructures and ecosystems
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs

Abstract

Early Classification of Time Series (ECTS) has been recognized as an important problem in many areas where decisions have to be taken as soon as possible, before the full data availability, while time pressure increases. Numerous ECTS approaches have been proposed, based on different triggering functions, each taking into account various pieces of information related to the incoming time series and/or the output of a classifier. Although their performances have been empirically compared in the literature, no studies have been carried out on the optimality of these triggering functions that involve ``man-tailored'' decision rules. Based on the same information, could there be better triggering functions? This paper presents one way to investigate this question by showing first how to translate ECTS problems into Reinforcement Learning (RL) ones, where the very same information is used in the state space. A thorough comparison of the performance obtained by ``handmade'' approaches and their ``RL-based'' counterparts has been carried out. A second question investigated in this paper is whether a different combination of information, defining the state space in RL systems, can achieve even better performance. Experiments show that the system we describe, called extsc{Alert}, significantly outperforms its state-of-the-art competitors on a large number of datasets.

Problem

• Optimizing triggering functions for early time series classification
• Comparing handmade and RL-based ECTS approaches
• Exploring state space combinations for better RL performance

Innovation

• Deep Reinforcement Learning techniques
• Optimized triggering function design
• Enhanced state space information combination