Delivering precise point and distributional forecasts across a spectrum of prediction horizons represents a significant and enduring challenge in the application of time-series forecasting within various industries.Prior research on developing deep learning models for time-series forecasting has often concentrated on isolated aspects, such as long-term point forecasting or short-term probabilistic estimations. This narrow focus may result in skewed methodological choices and hinder the adaptability of these models to uncharted scenarios.While there is a rising trend in developing universal forecasting models, a thorough understanding of their advantages and drawbacks, especially regarding essential forecasting needs like point and distributional forecasts across short and long horizons, is still lacking.In this paper, we present ProbTS, a benchmark tool designed as a unified platform to evaluate these fundamental forecasting needs and to conduct a rigorous comparative analysis of numerous cutting-edge studies from recent years.We dissect the distinctive data characteristics arising from disparate forecasting requirements and elucidate how these characteristics can skew methodological preferences in typical research trajectories, which often fail to fully accommodate essential forecasting needs.Building on this, we examine the latest models for universal time-series forecasting and discover that our analyses of methodological strengths and weaknesses are also applicable to these universal models.Finally, we outline the limitations inherent in current research and underscore several avenues for future exploration.

Numerous industrial sectors necessitate models capable of providing robust forecasts across various horizons. Despite the recent strides in crafting specific architectures for time-series forecasting and developing pre-trained universal models, a comprehensive examination of their capability in accommodating varied-horizon forecasting during inference is still lacking. This paper bridges this gap through the design and evaluation of the Elastic Time-Series Transformer (ElasTST). The ElasTST model incorporates a non-autoregressive design with placeholders and structured self-attention masks, warranting future outputs that are invariant to adjustments in inference horizons. A tunable version of rotary position embedding is also integrated into ElasTST to capture time-series-specific periods and enhance adaptability to different horizons. Additionally, ElasTST employs a multi-scale patch design, effectively integrating both fine-grained and coarse-grained information. During the training phase, ElasTST uses a horizon reweighting strategy that approximates the effect of random sampling across multiple horizons with a single fixed horizon setting. Through comprehensive experiments and comparisons with state-of-the-art time-series architectures and contemporary foundation models, we demonstrate the efficacy of ElasTST's unique design elements. Our findings position ElasTST as a robust solution for the practical necessity of varied-horizon forecasting.

Anomaly detection, particularly the detection of anomalous behaviors in dynamic and heterogeneous social networks, is becoming more and more crucial in real life. Traditional rule-based and feature-based methods cannot well capture the structural and temporal patterns of ever-changing user behaviors. Moreover, most of the existing works based on network embedding either rely on discretized snapshots, which have ignored accurate temporal relations among user behaviors and weakened the impact of new edges, or fail to utilize dynamic and heterogeneous information simultaneously to distinguish varying effects of new edges on existing nodes. In this paper, we propose an end-to-end continuous-time model, named Temporal Heterogeneous Graph Neural Network (THGNN), to detect anomalous behaviors (edges) in dynamic heterogeneous social networks. Specifically, the model constantly updates node embeddings by propagating the information of a new edge to its source and target nodes as well as their neighbors. In this process, heterogeneous encoders are employed to handle different types of nodes and edges. What is more, a novel dual-level distributive attention mechanism is designed to allocate the influence degree of a currently interacting node to its multiple neighbors, considering the combined effect of edge type and time interval information. That can be regarded as an extension of the classical aggregative attention mechanism in the opposite direction. Extensive experiments on four real-world datasets demonstrate that THGNN outperforms all the baselines on the task of anomalous edge detection, achieving an average AUC gain of 6% across all datasets.

Irregularly sampled multivariate time series are ubiquitous in various fields, particularly in healthcare, and exhibit two key characteristics: intra-series irregularity and inter-series discrepancy. Intra-series irregularity refers to the fact that time-series signals are often recorded at irregular intervals, while inter-series discrepancy refers to the significant variability in sampling rates among diverse series. However, recent advances in irregular time series have primarily focused on addressing intra-series irregularity, overlooking the issue of inter-series discrepancy. To bridge this gap, we present Warpformer, a novel approach that fully considers these two characteristics. In a nutshell, Warpformer has several crucial designs, including a specific input representation that explicitly characterizes both intra-series irregularity and inter-series discrepancy, a warping module that adaptively unifies irregular time series in a given scale, and a customized attention module for representation learning. Additionally, we stack multiple warping and attention modules to learn at different scales, producing multi-scale representations that balance coarse-grained and fine-grained signals for downstream tasks. We conduct extensive experiments on widely used datasets and a new large-scale benchmark built from clinical databases. The results demonstrate the superiority of Warpformer over existing state-of-the-art approaches.

Relation classification (RC) is an important task in knowledge extraction from texts, while data-driven approaches, although achieving high performance, heavily rely on a large amount of annotated training data. Recently, many few-shot RC models have been proposed and yielded promising results in general domain datasets, but when adapting to a specific domain, such as medicine, the performance drops dramatically. In this paper, we propose a Knowledge-Enhanced Few-shot RC model for the Domain Adaptation task (KEFDA), which incorporates general and domain-specific knowledge graphs (KGs) to the RC model to improve its domain adaptability. With the help of concept-level KGs, the model can better understand the semantics of texts and easily summarize the global semantics of relation types from only a few instances. To be more important, as a kind of meta-information, the manner of utilizing KGs can be transferred from existing tasks to new tasks, even across domains. Specifically, we design a knowledge-enhanced prototypical network to conduct instance matching, and a relation-meta learning network for implicit relation matching. The two scoring functions are combined to infer the relation type of a new instance. Experimental results on the Domain Adaptation Challenge in the FewRel 2.0 benchmark demonstrate that our approach significantly outperforms the state-of-the-art models (by 6.63% on average).