AdaTS: Learning Adaptive Time Series Representations via Dynamic Soft Contrasts

Denizhan Kara, Tomoyoshi Kimura, Jinyang Li, Bowen He, Yizhuo Chen, Yigong Hu, Hongjue Zhao, Shengzhong Liu, Tarek F. Abdelzaher

University of Illinois Urbana-Champaign· Shanghai Jiao Tong University

Time Series Representation Learning Contrastive Learning Self-Supervised Learning Adaptive Learning Dynamic Temporal Modeling

⋅ NeurIPS ⋅ OpenReview

Abstract

Learning robust representations from unlabeled time series is crucial, and contrastive learning offers a promising avenue. However, existing contrastive learning approaches for time series often struggle with defining meaningful similarities, tending to overlook inherent physical correlations and diverse, sequence-varying non-stationarity. This limits their representational quality and real-world adaptability.

To address these limitations, we introduce AdaTS, a novel adaptive soft contrastive learning strategy. AdaTS offers a compute-efficient solution centered on dynamic instance-wise and temporal assignments to enhance time series representations, specifically by:

leveraging Time-Frequency Coherence for robust physics-guided similarity measurement;
preserving relative instance similarities through ordinal consistency learning; and
dynamically adapting to sequence-specific non-stationarity with dynamic temporal assignments.

AdaTS is designed as a pluggable module to standard contrastive frameworks, achieving up to 13.7% accuracy improvements across diverse time series datasets and three state-of-the-art contrastive frameworks while enhancing robustness against label scarcity. The code will be publicly available upon acceptance.