In an era where machine learning permeates every facet of human existence, and data evolves incessantly, the application of machine learning models transcends mere data processing. It involves navigating constant changes exemplified by the phenomenon of concept drift, which often affects model performance. These drifts can be recurrent due to the cyclic nature of the underlying data generation processes, which could be influenced by recurrent phenomena such as weather and time of the day. Stream Learning on data streams with recurrent concept drifts attempts to learn from such streams of data. The survey underscores the significance of the field and its practical applications, delving into nuanced definitions of machine learning for data streams afflicted by recurrent concept drifts. It explores diverse methodological approaches, elucidating their key design components. Additionally, it examines various evaluation techniques, benchmark datasets, and available software tailored for simulating and analysing data streams with recurrent concept drifts. Concluding, the survey offers insights into potential avenues for future research in the field.
论文检索
输入标题、作者或关键词，从 2,101 篇学术成果中精准定位
Early predicting the onset of a disease is critical to timely and accurate clinical decision-making, where a model determines whether a patient will develop the disease n hours later. While deep learning algorithms have demonstrated great success using multivariate irregular time-series data such as electronic health records (EHRs), they often lack temporal robustness due to data scarcity problems becoming more prominent at multilevel as n increases. At event-level, the decreasing number of available events per trajectory increases uncertainty in anticipating future disease behavior. At trajectory-level, the scarcity of patient trajectories limits diversity in the training population, hindering the model's generalization. This work introduces Multi-TA, a multilevel temporal augmentation framework that leverages BERT-based temporal EHRs representation learning and a unified data augmentation approach, effectively addressing data scarcity issues at both event and trajectory levels. Validated on two real-world EHRs for septic shock, Multi-TA outperforms mixup and GAN-based state-of-the-art models across eight prediction windows, demonstrating improved temporal robustness. Further, we provide in-depth analyses on data augmentation.
Neural simulators for modeling complex dynamical systems have been extensively studied for various real-world applications, such as weather forecasting, ocean current prediction, and computational fluid dynamics simulation. Although they have demonstrated powerful fitting and predicting, most existing models are only built to learn single-system dynamics. Several advanced researches have considered learning dynamics across environments, which can exploit the potential commonalities among the dynamics across environments and adapt to new environments. However, these methods still are prone to scarcity problems where per-environment data is sparse or limited. Therefore, we propose a novel CoNDP (Context-Informed Neural ODE Processes) to achieve learning system dynamics from sparse observations across environments. It can fully use contextual information of each environment to better capture the intrinsic commonalities across environments and distinguishable differences among environments while modeling uncertainty of system evolution, producing more accurate predictions. Intensive experiments are conducted on five complex dynamical systems in various fields. Results show that the proposed CoNDP can achieve optimal results compared with common neural simulators and state-of-the-art cross-environmental models.
Sleep stage classification is crucial for sleep quality assessment and disease diagnosis. Although some recent studies have made great strides in sleep stage classification performance, direct application to multi-modal sleep data with cross-domain distributional variations still poses challenges: 1) How to retain the sleep knowledge acquired by the model from the source domain during cross-domain adaptation to avoid catastrophic forgetting. 2) How to evaluate the contribution of different modalities in identifying specific sleep stages to serve test-time adaptation (TTA). 3) How to dynamically adapt the sleep model to different distribution shift in data domains of different subjects. To address these challenges, we propose an Adaptive Test-Time Adaptation (ATTA) method, a multi-modal test-time adaptation method for sleep stage classification. Specifically, the intra-modal retained-adaptive module is proposed for adapting to the target domain data while retaining the sleep knowledge acquired from the source domain to avoid catastrophic forgetting. The inter-modal contribution assessment module is designed to adaptively assess the contribution of each modality to the identification of specific sleep stages. Furthermore, the adaptive learning rate strategy utilizes a memory bank to record data from different subjects during testing, and based on this, it measures the differences between the target subject and those in the memory bank. According to the difference, the model adapts to the subject samples with different learning rates. We conduct experiments on mutual migration on two sleep datasets, SleepEDF and SHHS. The results show that our ATTA method outperforms state-of-the-art baselines in sleep stage classification.
SDformer: Transformer with Spectral Filter and Dynamic Attention for Multivariate Time Series Long-term Forecasting
PDF ↗Transformer has gained widespread adoption in modeling time series due to the exceptional ability of its self-attention mechanism in capturing long-range dependencies. However, when processing time series data with numerous variates, the vanilla self-attention mechanism tends to distribute attention weights evenly and smoothly, causing row-homogenization in attention maps and further hampering time series forecasting. To tackle this issue, we propose an advanced Transformer architecture entitled SDformer, which designs two novel modules, Spectral-Filter-Transform (SFT) and Dynamic-Directional-Attention (DDA), and integrates them into the encoder of Transformer to achieve more intensive attention allocation. Specifically, the SFT module utilizes the Fast Fourier Transform to select the most prominent frequencies, along with a Hamming Window to smooth and denoise the filtered series data; The DDA module applies a specialized kernel function to the query and key vectors projected from the denoised data, concentrating this innovative attention mechanism more effectively on the most informative variates to obtain a sharper attention distribution. These two modules jointly enable attention weights to be more salient among numerous variates, which in turn enhances the attention's ability to capture multivariate correlations, improving the performance in forecasting. Extensive experiments on public datasets demonstrate its superior performance over other state-of-the-art models. Code is available at https://github.com/zhouziyu02/SDformer.
Time series self-supervised learning (SSL) aims to exploit unlabeled data for pre-training to mitigate the reliance on labels. Despite the great success in recent years, there is limited discussion on the potential noise in the time series, which can severely impair the performance of existing SSL methods. To mitigate the noise, the de facto strategy is to apply conventional denoising methods before model training. However, this pre-processing approach may not fully eliminate the effect of noise in SSL for two reasons: (i) the diverse types of noise in time series make it difficult to automatically determine suitable denoising methods; (ii) noise can be amplified after mapping raw data into latent space. In this paper, we propose denoising-aware contrastive learning (DECL), which uses contrastive learning objectives to mitigate the noise in the representation and automatically selects suitable denoising methods for every sample. Extensive experiments on various datasets verify the effectiveness of our method. The code is open-sourced.
Time series forecasting has essential applications across various domains. For instance, forecasting power time series can optimize energy usage and bolster grid stability and reliability. Existing models based on transformer architecture are limited to classical design, ignoring the impact of spatial information and noise on model architecture design. Therefore, we propose an atypical design of Transformer-based models for multivariate time series forecasting. This design consists of two critical components: (i) spectral clustering center of time series employed as the focal point for attention computation; (ii) alternating attention mechanism wherein each query transformer is compatible with spectral clustering centers, executing attention at the sequence level instead of the token level. The alternating design has a two-fold benefit: firstly, it eliminates the uncertainty noise present in the dependent variable sequence of the channel input, and secondly, it incorporates the Euclidean distance to mitigate the impact of extreme values on the attention matrix, thereby aligning predictions more closely to the sequence's natural progression. Experiments on ten real-world datasets, encompassing Wind, Electricity, Weather, and others, demonstrate that our Spectral Central Alternating Transformer (SCAT) outperforms state-of-the-art methods (SOTA) by an average of 42.8% in prediction performance in power time series forecasting.
NanoAdapt: Mitigating Negative Transfer in Test Time Adaptation with Extremely Small Batch Sizes
PDF ↗Test Time Adaptation (TTA) has garnered significant attention in recent years, with the research focus on addressing distribution shifts during test time. As one fundamental component of many TTA methods, the Batch Normalization (BN) layer plays a crucial role in enabling the model adaptability. However, existing BN strategies can prove detrimental when the batch size is (extremely) small. In numerous real-world scenarios, limited hardware resources or just-in-time demand often necessitates adjusting models with very small batch sizes, making existing methods less practical. In this paper, we first showcase and thoroughly analyze the negative transfer phenomenon in previous TTA methods encountering extremely small batch sizes. Subsequently, we propose a novel batch size-agnostic method called NanoAdapt to effectively mitigate the negative transfer even with batch size 1. NanoAdapt is composed of three key components: a dynamic BN calibration strategy that leverages historical information and the Taylor series to refine the statistics estimations, an entropy-weighted gradient accumulation strategy that uses the entropy of each sample's label prediction to weigh and accumulate the loss for backpropagation, and a novel proxy computation graph to capture the sample interactions. Extensive experiments are conducted to validate the superiority of NanoAdapt, showing its consistent efficacy in improving existing TTA methods.
Recent studies have raised questions about the suitability of the Transformer architecture for long sequence time-series forecasting. These forecasting models leverage Transformers to capture dependencies between multiple time steps in a time series, with embedding tokens composed of data from individual time steps. However, challenges arise when applying Transformers to predict long sequences with strong periodicity, leading to performance degradation and increased computational burden. Furthermore, embedding tokens formed one time step at a time may struggle to reveal meaningful information in long sequences, failing to capture correlations between different time steps. In this study, we propose Skip-Timeformer, a Transformer-based model that utilizes a skip-time interaction for long sequence time-series forecasting. Specifically, we decompose the time series into multiple subsequences based on different time intervals, embedding various time steps into variable tokens across multiple sequences. The skip-time interaction mechanism utilizes these variable tokens to capture dependencies in the skip-time dimension. Additionally, skip-time interaction is employed to learn dependencies between sequences missed by multiple skip time steps. The Skip-Timeformer model demonstrates state-of-the-art performance on various real-world datasets, further enhancing the long sequence forecasting capabilities of the Transformer variations and better adapting to arbitrary lookback windows.
VCformer: Variable Correlation Transformer with Inherent Lagged Correlation for Multivariate Time Series Forecasting
PDF ↗Multivariate time series (MTS) forecasting has been extensively applied across diverse domains, such as weather prediction and energy consumption. However, current studies still rely on the vanilla point-wise self-attention mechanism to capture cross-variable dependencies, which is inadequate in extracting the intricate cross-correlation implied between variables. To fill this gap, we propose Variable Correlation Transformer (VCformer), which utilizes Variable Correlation Attention (VCA) module to mine the correlations among variables. Specifically, based on the stochastic process theory, VCA calculates and integrates the cross-correlation scores corresponding to different lags between queries and keys, thereby enhancing its ability to uncover multivariate relationships. Additionally, inspired by Koopman dynamics theory, we also develop Koopman Temporal Detector (KTD) to better address non-stationarity in time series. The two key components enable VCformer to extract both multivariate correlations and temporal dependencies. Our extensive experiments on eight real-world datasets demonstrate the effectiveness of VCformer, achieving top-tier performance compared to other state-of-the-art baseline models. Code is available at this repository: https://github.com/CSyyn/VCformer.
Forecasting time series in imbalanced data presents a significant research challenge that requires considerable attention. Although there are specialized techniques available to tackle imbalanced time series prediction, existing approaches tend to prioritize extreme predictions at the expense of compromising the forecasting accuracy of normal samples. We in this paper propose an extreme penalized loss function that relaxes the constraint on overestimating extreme events, thereby imposing great penalties on both normal and underestimating extreme events. In addition, we provide a self-adaptive way for setting the hyperparameters of the loss function. Then, both the proposed loss function and an attention module are integrated with LSTM networks in a decomposition-based framework. Extensive experiments conducted on real-world datasets demonstrate the superiority of our framework compared to other state-of-the-art approaches for both time series prediction and block maxima prediction tasks.
Time series anomaly detection is a critical task with applications in various domains. Due to annotation challenges, self-supervised methods have become the mainstream approach for time series anomaly detection in recent years. However, current contrastive methods categorize data perturbations into binary classes, normal or anomaly, which lack clarity on the specific impact of different perturbation methods. Inspired by the hypothesis that "the higher the probability of misclassifying perturbation types, the higher the probability of anomalies", we propose PCRTA, our approach firstly devises a perturbation classifier to learn the pseudo-labels of data perturbations. Furthermore, for addressing "class collapse issue" in contrastive learning, we propose a perturbation guiding positive and negative samples selection strategy by introducing learnable perturbation classification networks. Extensive experiments on six realworld datasets demonstrate the significant superiority of our model over thirteen state-of-the-art competitors, and obtains average 5.14%, 8.24% improvement in F1 score and AUC-PR, respectively.
Exploiting self-supervised learning (SSL) to extract the universal representations of time series could not only capture the natural properties of time series but also offer huge help to the downstream tasks. Nevertheless, existing time series representation learning (TSRL) methods face challenges in attaining universality. Indeed, existing methods relying solely on one SSL strategy (either contrastive learning (CL) or generative) often fall short in capturing rich semantic information for various downstream tasks. Moreover, time series exhibit diverse distributions and inherent characteristics, particularly with the common occurrence of missing values, posing a notable challenge for existing backbones in effectively handling such diverse time series data. To bridge these gaps, we propose CTRL, a framework for universal TSRL. For the first time, we employ Neural Controlled Differential Equation (NCDE) as the backbone for TSRL, which captures the continuous processes and exhibits robustness to missing data. Additionally, a dual-task SSL strategy, integrating both reconstruction and contrasting tasks, is proposed to enrich the semantic information of the learned representations. Furthermore, novel hard negative construction and false negative elimination mechanisms are proposed to improve sampling efficiency and reduce sampling bias in CL. Finally, extensive experiments demonstrate the superiority of CTRL in forecasting, classification, and imputation tasks, particularly its outstanding robustness to missing data.
Recognizing domain generalization as a commonplace challenge in machine learning, data distribution might progressively evolve across a continuum of sequential domains in practical scenarios. While current methodologies primarily concentrate on bolstering model effectiveness within these new domains, they tend to neglect issues of fairness throughout the learning process. In response, we propose an innovative framework known as Disentanglement for Counterfactual Fairness-aware Domain Generalization (DCFDG). This approach adeptly removes domain-specific information and sensitive information from the embedded representation of classification features. To scrutinize the intricate interplay between semantic information, domain-specific information, and sensitive attributes, we systematically partition the exogenous factors into four latent variables. By incorporating fairness regularization, we utilize semantic information exclusively for classification purposes. Empirical validation on synthetic and authentic datasets substantiates the efficacy of our approach, demonstrating elevated accuracy levels while ensuring the preservation of fairness amidst the evolving landscape of continuous domains.
Recent sketch synthesis methods have demonstrated the capability of generating lifelike outcomes. However, these methods directly encode the entire sketches making it challenging to decouple the strokes from the sketches and have difficulty in controlling local sketch synthesis, e.g., stroke editing. Besides, the sketch editing task encounters the issue of accurately positioning the edited strokes, because users may not be able to draw on the exact position and the same stroke may appear in various locations in different sketches. We propose SketchEdit to realize flexible editing of sketches at the stroke-level for the first time. To tackle the challenge of decoupling strokes, SketchEdit divides a drawing sequence of a sketch into a series of strokes based on the pen state, aligns the stroke segments to have the same starting position, and learns the embeddings of every stroke by a proposed stroke encoder. Moreover, we overcome the problem of stroke placement via a diffusion process, which progressively generates the locations for the strokes to be synthesized, using the stroke features as the guiding condition. Experiments demonstrate that SketchEdit is effective for stroke-level sketch editing and sketch reconstruction. The source code is publicly available at https://github.com/CMACH508/SketchEdit/.
Wearable sensors play a crucial role in real-world scenarios, such as human activity recognition, sleep monitoring and electrocardiogram monitoring. However, deploying classifiers on them is challenged by distribution shifts across users and devices. Unsupervised domain adaptation (UDA) is proposed to address this, yet existing methods mostly focus on feature distribution shift, neglecting the potential misclassification due to label shift. In this paper, we propose Domain adaptation under label shift for Wearable sensor with Learnable Reweighting (DWLR) to handle both feature and label shifts. Specifically, DWLR employs learnable reweighting to align label distributions between source and target domains. It incorporates elements of information gain during the reweighting process to counter potential distribution shift that could emerge from over-reliance on data with high-confidence pseudo labels. Importantly, since wearable sensor data is time-series data, and can be subjected to distribution shifts originating from either the time domain, the frequency domain, or both, DWLR performs reweighting and alignment separately in these two domains to more robustly handle potential feature distribution shifts. Extensive experiments on three distinct wearable sensor datasets demonstrate the effectiveness of DWLR, yielding a remarkable average performance improvement of 5.85%.
Negative-Binomial Randomized Gamma Dynamical Systems for Heterogeneous Overdispersed Count Time Sequences
PDF ↗Modeling count-valued time sequences has been receiving growing interests because count time sequences naturally arise in physical and social domains. Poisson gamma dynamical systems (PGDSs) are newly-developed methods, which can well capture the expressive latent transition structure and bursty dynamics behind count sequences. In particular, PGDSs demonstrate superior performance in terms of data imputation and prediction, compared with canonical linear dynamical system (LDS) based methods. Despite these advantages, PGDS cannot capture the heterogeneous overdispersed behaviours of the underlying dynamic processes. To mitigate this defect, we propose a negative-binomial-randomized gamma Markov process, which not only significantly improves the predictive performance of the proposed dynamical system, but also facilitates the fast convergence of the inference algorithm. Moreover, we develop methods to estimate both factor-structured and graph-structured transition dynamics, which enable us to infer more explainable latent structure, compared with PGDSs. Finally, we demonstrate the explainable latent structure learned by the proposed method, and show its superior performance in imputing missing data and forecasting future observations, compared with the related models.
Time series forecasting (TSF) plays a pivotal role in many real-world applications. Recently, the utilization of Large Language Models (LLM) in TSF has demonstrated exceptional predictive performance, surpassing most task-specific forecasting models. The success of LLM-based forecasting methods underscores the importance of causal dependence modeling and pre-trained knowledge transfer. However, challenges persist in directly applying LLM to TSF, i.e., the unacceptable parameter scales for resource-intensive model optimization, and the significant gap of feature space between structural numerical time series and natural language. To this end, we propose LeRet, a Language-empowered Retentive network for TSF. Technically, inspired by the causal extraction in LLM, we propose a causal dependence learner, enhanced by a patch-level pre-training task, to capture sequential causal evolution. To minimize the gap between numeric and language, we initialize a language description protocol for time series and design a TS-related language knowledge extractor to learn from language description, avoiding training with large-scale parameters. Finally, we dedicatedly achieve a Language-TS Modality Integrator for the fusion of two types data, and enable language-empowered sequence forecasting. Extensive evaluations demonstrate the effectiveness of our LeRet, especially reveal superiority on few-shot, and zero-shot forecasting tasks. Code is available at https://github.com/hqh0728/LeRet.
Modern graph representation learning works mostly under the assumption of dealing with regularly sampled temporal graph snapshots, which is far from realistic, e.g., social networks and physical systems are characterized by continuous dynamics and sporadic observations. To address this limitation, we introduce the Temporal Graph Ordinary Differential Equation (TG-ODE) framework, which learns both the temporal and spatial dynamics from graph streams where the intervals between observations are not regularly spaced. We empirically validate the proposed approach on several graph benchmarks, showing that TG-ODE can achieve state-of-the-art performance in irregular graph stream tasks.
Spatiotemporal forecasting techniques are significant for various domains such as transportation, energy, and weather. Accurate prediction of spatiotemporal series remains challenging due to the complex spatiotemporal heterogeneity. In particular, current end-to-end models are limited by input length and thus often fall into spatiotemporal mirage, i.e., similar input time series followed by dissimilar future values and vice versa. To address these problems, we propose a novel self-supervised pre-training framework Spatial-Temporal-Decoupled Masked Pre-training (STD-MAE) that employs two decoupled masked autoencoders to reconstruct spatiotemporal series along the spatial and temporal dimensions. Rich-context representations learned through such reconstruction could be seamlessly integrated by downstream predictors with arbitrary architectures to augment their performances. A series of quantitative and qualitative evaluations on four widely used benchmarks (PEMS03, PEMS04, PEMS07, and PEMS08) are conducted to validate the state-of-the-art performance of STD-MAE. Codes are available at https://github.com/Jimmy-7664/STD-MAE.