FIC-TSC: Learning Time Series Classification with Fisher Information Constraint

2025-05-09

AI Summary

Distribution shift (domain shift) between training and test sets severely degrades generalization in time-series classification. To address this, we propose the first Fisher information-constrained training framework specifically designed for time-series classification. Theoretically, we prove that our method guides optimization toward flatter minima in the loss landscape, thereby enhancing cross-domain robustness. Methodologically, we incorporate Fisher information as an explicit regularizer into the loss function, coupled with differentiable optimization constraints, temporal feature normalization adaptation, and a multi-scale convolutional-attention fusion architecture. Extensive evaluation on 30 UEA multivariate and 85 UCR univariate benchmark datasets demonstrates that our approach consistently outperforms 14 state-of-the-art methods, achieving average accuracy improvements of 1.8–3.2 percentage points and significantly mitigating performance degradation induced by distribution shift.

Technology Category

• Machine Learning: Time-Series/Data Streams
• Constraint Satisfaction and Optimization: Distributed CSP/Optimization
• Computer Vision: Adversarial Attacks & Robustness

Application Category

• Search and Retrieval-Augmented AI: Vertical and domain-specific search
• Graph Algorithms and Modeling for the Web: Algorithms and analysis for heterogeneous, signed, attributed, multi-relational, temporal, higher-order, and annotated Web-related graphs
• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods

Abstract

Analyzing time series data is crucial to a wide spectrum of applications, including economics, online marketplaces, and human healthcare. In particular, time series classification plays an indispensable role in segmenting different phases in stock markets, predicting customer behavior, and classifying worker actions and engagement levels. These aspects contribute significantly to the advancement of automated decision-making and system optimization in real-world applications. However, there is a large consensus that time series data often suffers from domain shifts between training and test sets, which dramatically degrades the classification performance. Despite the success of (reversible) instance normalization in handling the domain shifts for time series regression tasks, its performance in classification is unsatisfactory. In this paper, we propose extit{FIC-TSC}, a training framework for time series classification that leverages Fisher information as the constraint. We theoretically and empirically show this is an efficient and effective solution to guide the model converge toward flatter minima, which enhances its generalizability to distribution shifts. We rigorously evaluate our method on 30 UEA multivariate and 85 UCR univariate datasets. Our empirical results demonstrate the superiority of the proposed method over 14 recent state-of-the-art methods.

Problem

• Addressing domain shifts in time series classification tasks
• Improving model generalizability to distribution shifts
• Enhancing classification performance with Fisher information constraints

Innovation

• Leverages Fisher information as constraint
• Guides model to converge toward flatter minima
• Enhances generalizability to distribution shifts