Foundation models pre-trained on large-scale source datasets are reshaping the traditional training paradigm for time series classification. However, existing time series foundation models primarily focus on forecasting tasks and often overlook classification-specific challenges, such as modeling interpretable shapelets that capture class-discriminative temporal features. To bridge this gap, we propose UniShape, a unified shape-aware foundation model designed for time series classification. UniShape incorporates a shape-aware adapter that adaptively aggregates multiscale discriminative subsequences (shapes) into class tokens, effectively selecting the most relevant subsequence scales to enhance model interpretability. Meanwhile, a prototype-based pretraining module is introduced to jointly learn instance- and shape-level representations, enabling the capture of transferable shape patterns. Pre-trained on a large-scale multi-domain time series dataset comprising 1.89 million samples, UniShape exhibits superior generalization across diverse target domains. Experiments on 128 UCR datasets and 30 additional time series datasets demonstrate that UniShape achieves state-of-the-art classification performance, with interpretability and ablation analyses further validating its effectiveness.

Time series domain adaptation aims to transfer the complex temporal dependence from the labeled source domain to the unlabeled target domain. Recent advances leverage the stable causal mechanism over observed variables to model the domain-invariant temporal dependence. However, modeling precise causal structures in high-dimensional data, such as videos, remains challenging. Additionally, direct causal edges may not exist among observed variables (e.g., pixels). These limitations hinder the applicability of existing approaches to real-world scenarios. To address these challenges, we find that the high-dimension time series data are generated from the low-dimension latent variables, which motivates us to model the causal mechanisms of the temporal latent process. Based on this intuition, we propose a latent causal mechanism identification framework that guarantees the uniqueness of the reconstructed latent causal structures. Specifically, we first identify latent variables by utilizing sufficient changes in historical information. Moreover, by enforcing the sparsity of the relationships of latent variables, we can achieve identifiable latent causal structures. Built on the theoretical results, we develop the Latent Causality Alignment (LCA) model that leverages variational inference, which incorporates an intra-domain latent sparsity constraint for latent structure reconstruction and an inter-domain latent sparsity constraint for domain-invariant structure reconstruction. Experiment results on eight benchmarks show a general improvement in the domain-adaptive time series classification and forecasting tasks, highlighting the effectiveness of our method in real-world scenarios.

Source-free domain adaptation (SFDA) adapts a pre-trained model from a labeled source domain to an unlabeled target domain without source data access, preserving privacy. While SFDA is common in computer vision, it remains largely unexplored in time series analysis, where existing methods struggle to capture temporal dynamics and often produce overconfident predictions on out-of-distribution samples. We propose MAsk And imPUte (MAPU), which tackles temporal consistency through a novel imputation task, where randomly masked time series signals are recovered within the learned embedding space. During adaptation, a dedicated temporal imputer guides the target model to generate features that maintain temporal consistency with source features. However, MAPU relies on standard softmax predictions, leading to overconfident predictions on target samples that fall outside the source domain's support. To address this limitation, we introduce Evidential-MAPU (E-MAPU), which leverages evidential uncertainty estimation to identify these out-of-support samples and adapts the feature extractor to map them closer to the source domain's support, while maintaining the classifier fixed. Extensive experiments on five real-world time series datasets demonstrate significant performance improvements over existing methods. Our approaches effectively handle various time series domain adaptation challenges while maintaining computational efficiency, achieving state-of-the-art performance through its uncertainty-aware adaptation strategy.

Deep learning has significantly advanced time series forecasting through its powerful capacity to capture sequence relationships. However, training these models with the Mean Square Error (MSE) loss often results in over-smooth predictions, making it challenging to handle the complexity and learn high-entropy features from time series data with high variability and unpredictability. In this work, we introduce a novel approach by tokenizing time series values to train forecasting models via cross-entropy loss, while considering the continuous nature of time series data. Specifically, we propose a Hierarchical Classification Auxiliary Network, HCAN, a general model-agnostic component that can be integrated with any forecasting model. HCAN is based on a Hierarchy-Aware Attention module that integrates multi-granularity high-entropy features at different hierarchy levels. At each level, we assign a class label for timesteps to train an Uncertainty-Aware Classifier. This classifier mitigates the over-confidence in softmax loss via evidence theory. We also implement a Hierarchical Consistency Loss to maintain prediction consistency across hierarchy levels. Extensive experiments integrating HCAN with state-of-the-art forecasting models demonstrate substantial improvements over baselines on several real-world datasets.

Shapelets are discriminative subsequences (or shapes) with high interpretability in time series classification. Due to the time-intensive nature of shapelet discovery, existing shapelet-based methods mainly focus on selecting discriminative shapes while discarding others to achieve candidate subsequence sparsification. However, this approach may exclude beneficial shapes and overlook the varying contributions of shapelets to classification performance. To this end, we propose a Soft sparse Shapes (SoftShape) model for efficient time series classification. Our approach mainly introduces soft shape sparsification and soft shape learning blocks. The former transforms shapes into soft representations based on classification contribution scores, merging lower-scored ones into a single shape to retain and differentiate all subsequence information. The latter facilitates intra- and inter-shape temporal pattern learning, improving model efficiency by using sparsified soft shapes as inputs. Specifically, we employ a learnable router to activate a subset of class-specific expert networks for intra-shape pattern learning. Meanwhile, a shared expert network learns inter-shape patterns by converting sparsified shapes into sequences. Extensive experiments show that SoftShape outperforms state-of-the-art methods and produces interpretable results.

Estimating Individual Treatment Effects (ITE) from observational data is challenging due to covariate shift and counterfactual absence. While existing methods attempt to balance distributions globally, they often lack fine-grained sample-level alignment, especially in scenarios with significant individual heterogeneity. To address these issues, we reconsider counterfactual as a proxy to emulate balanced randomization. Furthermore, we derive a theoretical bound that links the expected ITE estimation error to both factual prediction errors and representation distances between factuals and counterfactuals. Building on this theoretical foundation, we propose FCCL, a novel method designed to effectively capture the nuances of potential outcomes under different treatments by (i) generating diffeomorphic counterfactuals that adhere to the data manifold while maintaining high semantic similarity to their factual counterparts, and (ii) mitigating distribution shift via sample-level alignment grounded in our derived generalization-error bound, which considers factual-counterfactual similarity and category consistency. Extensive evaluations on benchmark datasets demonstrate that FCCL outperforms 13 state-of-the-art methods, particularly in capturing individual-level heterogeneity and handling sparse boundary samples.

Time series forecasting, which aims to predict future values based on historical data, has garnered significant attention due to its broad range of applications. However, real-world time series often exhibit heterogeneous pattern evolution across segments, such as seasonal variations, regime changes, or contextual shifts, making accurate forecasting challenging. Existing approaches, which typically train a single model to capture all these diverse patterns, often struggle with the pattern drifts between patches and may lead to poor generalization. To address these challenges, we propose TFPS, a novel architecture that leverages pattern-specific experts for more accurate and adaptable time series forecasting. TFPS employs a dual-domain encoder to capture both time-domain and frequency-domain features, enabling a more comprehensive understanding of temporal dynamics. It then performs subspace clustering to dynamically identify distinct patterns across data segments. Finally, these patterns are modeled by specialized experts, allowing the model to learn multiple predictive functions. Extensive experiments on real-world datasets demonstrate that TFPS outperforms state-of-the-art methods, particularly on datasets exhibiting significant distribution shifts. The data and code are available: https://github.com/syrGitHub/TFPS.

Time series data, characterized by its intrinsic long and short-range dependencies, poses a unique challenge across analytical applications. While Transformer-based models excel at capturing long-range dependencies, they face limitations in noise sensitivity, computational efficiency, and overfitting with smaller datasets. In response, we introduce a novel **T**ime **S**eries **L**ightweight **A**daptive **Net**work (**TSLANet**), as a universal convolutional model for diverse time series tasks. Specifically, we propose an Adaptive Spectral Block, harnessing Fourier analysis to enhance feature representation and to capture both long-term and short-term interactions while mitigating noise via adaptive thresholding. Additionally, we introduce an Interactive Convolution Block and leverage self-supervised learning to refine the capacity of TSLANet for decoding complex temporal patterns and improve its robustness on different datasets. Our comprehensive experiments demonstrate that TSLANet outperforms state-of-the-art models in various tasks spanning classification, forecasting, and anomaly detection, showcasing its resilience and adaptability across a spectrum of noise levels and data sizes. The code is available at https://github.com/emadeldeen24/TSLANet.

Unsupervised domain adaptation (UDA) has successfully addressed the domain shift problem for visual applications. Yet, these approaches may have limited performance for time series data due to the following reasons. First, they mainly rely on the large-scale dataset (i.e., ImageNet) for source pretraining, which is not applicable for time series data. Second, they ignore the temporal dimension on the feature space of the source and target domains during the domain alignment step. Finally, most of the prior UDA methods can only align the global features without considering the fine-grained class distribution of the target domain. To address these limitations, we propose a SeLf-supervised AutoRegressive Domain Adaptation (SLARDA) framework. In particular, we first design a self-supervised (SL) learning module that uses forecasting as an auxiliary task to improve the transferability of source features. Second, we propose a novel autoregressive domain adaptation technique that incorporates temporal dependence of both source and target features during domain alignment. Finally, we develop an ensemble teacher model to align class-wise distribution in the target domain via a confident pseudo labeling approach. Extensive experiments have been conducted on three real-world time series applications with 30 cross-domain scenarios. The results demonstrate that our proposed SLARDA method significantly outperforms the state-of-the-art approaches for time series domain adaptation. Our source code is available at: https://github.com/mohamedr002/SLARDA.

Source-free domain adaptation (SFDA) aims to adapt a pretrained model from a labeled source domain to an unlabeled target domain without access to the source domain data, preserving source domain privacy. Despite its prevalence in visual applications, SFDA is largely unexplored in time series applications. The existing SFDA methods that are mainly designed for visual applications may fail to handle the temporal dynamics in time series, leading to impaired adaptation performance. To address this challenge, this paper presents a simple yet effective approach for source-free domain adaptation on time series data, namely MAsk and imPUte (MAPU). First, to capture temporal information of the source domain, our method performs random masking on the time series signals while leveraging a novel temporal imputer to recover the original signal from a masked version in the embedding space. Second, in the adaptation step, the imputer network is leveraged to guide the target model to produce target features that are temporally consistent with the source features. To this end, our MAPU can explicitly account for temporal dependency during the adaptation while avoiding the imputation in the noisy input space. Our method is the first to handle temporal consistency in SFDA for time series data and can be seamlessly equipped with other existing SFDA methods. Extensive experiments conducted on three real-world time series datasets demonstrate that our MAPU achieves significant performance gain over existing methods. Our code is available at: https://github.com/mohamedr002/MAPU_SFDA_TS.

Learning time-series representations when only unlabeled data or few labeled samples are available can be a challenging task. Recently, contrastive self-supervised learning has shown great improvement in extracting useful representations from unlabeled data via contrasting different augmented views of data. In this work, we propose a novel Time-Series representation learning framework via Temporal and Contextual Contrasting (TS-TCC) that learns representations from unlabeled data with contrastive learning. Specifically, we propose time-series-specific weak and strong augmentations and use their views to learn robust temporal relations in the proposed temporal contrasting module, besides learning discriminative representations by our proposed contextual contrasting module. Additionally, we conduct a systematic study of time-series data augmentation selection, which is a key part of contrastive learning. We also extend TS-TCC to the semi-supervised learning settings and propose a Class-Aware TS-TCC (CA-TCC) that benefits from the available few labeled data to further improve representations learned by TS-TCC. Specifically, we leverage the robust pseudo labels produced by TS-TCC to realize a class-aware contrastive loss. Extensive experiments show that the linear evaluation of the features learned by our proposed framework performs comparably with the fully supervised training. Additionally, our framework shows high efficiency in few labeled data and transfer learning scenarios.

Unsupervised domain adaptation methods aim at generalizing well on unlabeled test data that may have a different (shifted) distribution from the training data. Such methods are typically developed on image data, and their application to time series data is less explored. Existing works on time series domain adaptation suffer from inconsistencies in evaluation schemes, datasets, and backbone neural network architectures. Moreover, labeled target data are often used for model selection, which violates the fundamental assumption of unsupervised domain adaptation. To address these issues, we develop a benchmarking evaluation suite ( AdaTime ) to systematically and fairly evaluate different domain adaptation methods on time series data. Specifically, we standardize the backbone neural network architectures and benchmarking datasets, while also exploring more realistic model selection approaches that can work with no labeled data or just a few labeled samples. Our evaluation includes adapting state-of-the-art visual domain adaptation methods to time series data as well as the recent methods specifically developed for time series data. We conduct extensive experiments to evaluate 11 state-of-the-art methods on five representative datasets spanning 50 cross-domain scenarios. Our results suggest that with careful selection of hyper-parameters, visual domain adaptation methods are competitive with methods proposed for time series domain adaptation. In addition, we find that hyper-parameters could be selected based on realistic model selection approaches. Our work unveils practical insights for applying domain adaptation methods on time series data and builds a solid foundation for future works in the field. The code is available at github.com/emadeldeen24/AdaTime .

Learning decent representations from unlabeled time-series data with temporal dynamics is a very challenging task. In this paper, we propose an unsupervised Time-Series representation learning framework via Temporal and Contextual Contrasting (TS-TCC), to learn time-series representation from unlabeled data. First, the raw time-series data are transformed into two different yet correlated views by using weak and strong augmentations. Second, we propose a novel temporal contrasting module to learn robust temporal representations by designing a tough cross-view prediction task. Last, to further learn discriminative representations, we propose a contextual contrasting module built upon the contexts from the temporal contrasting module. It attempts to maximize the similarity among different contexts of the same sample while minimizing similarity among contexts of different samples. Experiments have been carried out on three real-world time-series datasets. The results manifest that training a linear classifier on top of the features learned by our proposed TS-TCC performs comparably with the supervised training. Additionally, our proposed TS-TCC shows high efficiency in few-labeled data and transfer learning scenarios. The code is publicly available at https://github.com/emadeldeen24/TS-TCC.