arXiv subjects

Chiao-Tung Chan

Publications and source records attributed to Chiao-Tung Chan.

3 recordsLinked to original sources

arXiv2026

Perseus: Interactive Time Series Segmentation with Sparse Supervision via Stateful Memory

Real-world systems, ranging from industrial manufacturing to wearable healthcare, generate multivariate time series with hierarchical states ranging from coarse regimes to fine-grained events. Unlike zero- or few-shot segmentation, our setting uses dense state labels for model training. Sparse expert prompts provide inference-time corrections that resolve sequence-specific ambiguities without retraining. In practice, this feedback is grouped around selected events or transitions, leaving large portions of the timeline unprompted. The prompt-based sliding-window baselines evaluated here are stateless with respect to user interaction history: they use guidance only within the current window and cannot retain it across these gaps. To address this, we propose Perseus (Persistent Segmentation with User Supervision), a framework that transitions from synchronous processing to asynchronous state management. Perseus decouples supervision from inference via a distinct Write-Read architecture: grouped user cues are asynchronously encoded into a persistent memory bank (Write), which is then actively queried by the inference engine (Read) to service unprompted windows. This mechanism bridges supervision gaps by conditioning predictions on a global history of interactions rather than solely on local inputs. Extensive experiments on six datasets demonstrate that while evaluated stateless prompting baselines degrade significantly under grouped supervision, Perseus maintains robustness and achieves up to 85% accuracy improvement in multi-granularity settings. Code and preprocessing instructions are available at https://github.com/blacksnail789521/Perseus.

cs.LG↗

arXiv2024

Self-Supervised Learning of Disentangled Representations for Multivariate Time-Series

Multivariate time-series data in fields like healthcare and industry are informative but challenging due to high dimensionality and lack of labels. Recent self-supervised learning methods excel in learning rich representations without labels but struggle with disentangled embeddings and inductive bias issues like transformation-invariance. To address these challenges, we introduce TimeDRL, a framework for multivariate time-series representation learning with dual-level disentangled embeddings. TimeDRL features: (i) disentangled timestamp-level and instance-level embeddings using a [CLS] token strategy; (ii) timestamp-predictive and instance-contrastive tasks for representation learning; and (iii) avoidance of augmentation methods to eliminate inductive biases. Experiments on forecasting and classification datasets show TimeDRL outperforms existing methods, with further validation in semi-supervised settings with limited labeled data.

cs.LG↗

arXiv2024

TimeDRL: Disentangled Representation Learning for Multivariate Time-Series

Multivariate time-series data in numerous real-world applications (e.g., healthcare and industry) are informative but challenging due to the lack of labels and high dimensionality. Recent studies in self-supervised learning have shown their potential in learning rich representations without relying on labels, yet they fall short in learning disentangled embeddings and addressing issues of inductive bias (e.g., transformation-invariance). To tackle these challenges, we propose TimeDRL, a generic multivariate time-series representation learning framework with disentangled dual-level embeddings. TimeDRL is characterized by three novel features: (i) disentangled derivation of timestamp-level and instance-level embeddings from patched time-series data using a [CLS] token strategy; (ii) utilization of timestamp-predictive and instance-contrastive tasks for disentangled representation learning, with the former optimizing timestamp-level embeddings with predictive loss, and the latter optimizing instance-level embeddings with contrastive loss; and (iii) avoidance of augmentation methods to eliminate inductive biases, such as transformation-invariance from cropping and masking. Comprehensive experiments on 6 time-series forecasting datasets and 5 time-series classification datasets have shown that TimeDRL consistently surpasses existing representation learning approaches, achieving an average improvement of forecasting by 58.02% in MSE and classification by 1.48% in accuracy. Furthermore, extensive ablation studies confirmed the relative contribution of each component in TimeDRL's architecture, and semi-supervised learning evaluations demonstrated its effectiveness in real-world scenarios, even with limited labeled data. The code is available at https://github.com/blacksnail789521/TimeDRL.

cs.LG↗