Learning Time-Series Representations by Hierarchical Uniformity-Tolerance Latent Balancing

2025-10-02

AI Summary

In time-series representation learning, existing methods struggle to simultaneously achieve embedding space uniformity and tolerance, leading to insufficient modeling of temporal dependencies. To address this, we propose a hierarchical contrastive learning framework. Its key contributions are: (1) a hierarchical angular margin loss that explicitly encodes geometric relationships among instance-level and time-level positive/negative samples; and (2) a temperature-scheduling mechanism that dynamically balances representation discriminability and robustness during training. Evaluated on 128 UCR and 30 UEA benchmark datasets, our method consistently outperforms state-of-the-art contrastive and supervised approaches. It achieves significant gains in classification accuracy and competitive performance in anomaly detection, empirically validating the effectiveness of jointly optimizing uniformity and tolerance through hierarchical contrastive learning for time-series representation.

Technology Category

• Machine Learning: Representation Learning
• Computer Vision: Representation Learning for Vision
• Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning

Application Category

• Graph Algorithms and Modeling for the Web: Graph embeddings and representation learning for Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

We propose TimeHUT, a novel method for learning time-series representations by hierarchical uniformity-tolerance balancing of contrastive representations. Our method uses two distinct losses to learn strong representations with the aim of striking an effective balance between uniformity and tolerance in the embedding space. First, TimeHUT uses a hierarchical setup to learn both instance-wise and temporal information from input time-series. Next, we integrate a temperature scheduler within the vanilla contrastive loss to balance the uniformity and tolerance characteristics of the embeddings. Additionally, a hierarchical angular margin loss enforces instance-wise and temporal contrast losses, creating geometric margins between positive and negative pairs of temporal sequences. This approach improves the coherence of positive pairs and their separation from the negatives, enhancing the capture of temporal dependencies within a time-series sample. We evaluate our approach on a wide range of tasks, namely 128 UCR and 30 UAE datasets for univariate and multivariate classification, as well as Yahoo and KPI datasets for anomaly detection. The results demonstrate that TimeHUT outperforms prior methods by considerable margins on classification, while obtaining competitive results for anomaly detection. Finally, detailed sensitivity and ablation studies are performed to evaluate different components and hyperparameters of our method.

Problem

• Balancing uniformity and tolerance in time-series contrastive representations
• Learning hierarchical instance-wise and temporal information from time-series
• Improving temporal dependency capture through geometric margin enforcement

Innovation

• Hierarchical uniformity-tolerance balancing for contrastive representations
• Temperature scheduler integrated into contrastive loss
• Hierarchical angular margin loss enforces geometric separations