With the wide adoption of online education platforms, adaptive learning systems have become increasingly important. Learning Path Recommendation (LPR) aims to dynamically adjust learning content to optimize learning efficiency based on individual student needs. However, current LPR methods suffer from sparse reward for precise assessment and only focus on anonymous sessions that overlook more personalized and effective paths. To address these challenges, we propose UNO, UNified Offline Training Paradigm for Learning Path Recommendation. This approach introduces an offline training paradigm in RL-based LPR to provide dense process rewards by a personalized advantage based on a reward model, which can estimate the students' internal knowledge levels on the learning targets. Additionally, we propose UniLPR model, a personalized recommendation system that unifies modeling the implicit relationships between students' long-term accumulation and evolving requirements for questions, and refines through Group Relative Policy Optimization(GRPO). Finally, we design learning tasks that encompass historical reviewing, recent learning, and long-term exploratory learning to simulate the comprehensive and diverse learning needs of students. Our UNO achieves state-of-the-art performance across all tasks, demonstrating its effectiveness.

Dynamic graph learning plays a pivotal role in modeling evolving relationships over time, especially for temporal link prediction tasks in domains such as traffic systems, social networks, and recommendation platforms. While Transformer-based models have demonstrated strong performance by capturing long-range temporal dependencies, their reliance on self-attention results in quadratic complexity with respect to sequence length, limiting scalability on high-frequency or large-scale graphs. In this work, we revisit the necessity of self-attention in dynamic graph modeling. Inspired by recent findings that attribute the success of Transformers more to their architectural design than attention itself, we propose GLFormer, a novel attention-free Transformer-style framework for dynamic graphs. GLFormer introduces an adaptive token mixer that performs context-aware local aggregation based on interaction order and time intervals. To capture long-term dependencies, we further design a hierarchical aggregation module that expands the temporal receptive field by stacking local token mixers across layers. Experiments on six widely used dynamic graph benchmarks show that GLFormer achieves competitive or superior performance, which reveals that attention-free architectures can match or surpass Transformer baselines in dynamic graph settings with significantly improved efficiency.

Recent years have witnessed the rapid development of deep-learning-based, graph-neural-network-based forecasting methods for modern intelligent transportation systems. However, most existing work focuses exclusively on capturing spatio-temporal dependencies from historical traffic data, while overlooking the fact that suddenly occurring transportation incidents, such as traffic accidents and adverse weather, serve as external disturbances that can substantially alter temporal patterns. We argue that this issue has become a major obstacle to modeling the dynamics of traffic systems and improving prediction accuracy, but the unpredictability of incidents makes it difficult to observe patterns from historical sequences. To address these challenges, this paper proposes a novel framework named the Incident-Guided Spatiotemporal Graph Neural Network (IGSTGNN). IGSTGNN explicitly models the incident's impact through two core components: an Incident-Context Spatial Fusion (ICSF) module to capture the initial heterogeneous spatial influence, and a Temporal Incident Impact Decay (TIID) module to model the subsequent dynamic dissipation. To facilitate research on the spatio-temporal impact of incidents on traffic flow, a large-scale dataset is constructed and released, featuring incident records that are time-aligned with traffic time series. On this new benchmark, the proposed IGSTGNN framework is demonstrated to achieve state-of-the-art performance. Furthermore, the generalizability of the ICSF and TIID modules is validated by integrating them into various existing models.

The combination of temporal graph neural networks (TGNNs) and relative positional features is effective for dynamic link prediction tasks because it improves the modeling of relationships between nodes. We summarize the positional feature as Temporal Relative Positional Encoding (TRPE), which incorporates local subgraph computation and temporal decay effects compared to RPE in static graph learning. However, existing dynamic graph learning methods with TRPE are limited by high computational costs and poor scalability due to information loss in neighborhood compression. To address this, we introduce a scalable framework that enhances TRPE computation by integrating it with temporal clusters. Our method replaces high-order relative position information with cluster-based computation to reduce computation costs. Each node is assigned to a cluster based on its structural role, rather than just randomly sketched, reducing information loss during neighborhood compression. TGNNs with our proposed framework are more expressive than existing dynamic graph learning methods and offer greater scalability than existing TRPE models. Our experimental results on seven standard temporal link prediction benchmarks demonstrate that our proposed model achieves comparable or superior performance compared to state-of-the-art models.

With the increasing collection of electronic health records (EHRs), deep learning has become a crucial tool for real-time treatment analysis. However, due to patient privacy concerns, the scarcity of labeled data limits the end-to-end models that rely on large training data. Self-supervised pretraining offers a promising solution. Nevertheless, applying pretraining to EHRs faces two key issues: (1) EHRs exhibit multimodality, including monitoring data and recorded clinical note. For multimodal pretraining, designing a self-supervised task that can establish cross-modal associations while preserving all modal-unique information remains challenging. (2) Both modalities are sequential and irregular, with varying intervals between monitoring or records. Aligning monitoring times with recorded times poses a significant issue for fine-grained cross-modal pretraining. Existing pretraining models either focus on a single modality or only models regular data, failing to address them together. To fill this gap and fully utilize unlabel EHR data, we propose a p retraining model to learn patient r epresentation using unlabel i rregular m ultimodal E HRs, named PRIME. We first utilize a multi-element encoding module to extract patient condition snapshots from both modalities. Then, to construct multiple aligned cross-modal positive sample pairs that span the entire treatment process from irregular data, we employ patient condition alignment modules that integrate time-aware and feature-aware components to transfer snapshots to the aligned timestamps. Next, to preserve both shared and unique information of each modality, our decoupled representation learning strategy first uses a constraint matrix to separate shared information. We then employ contrastive-based cross-modal learning and reconstruction-based intra-modal learning to model shared and complete information, respectively. Extensive experiments on two real-world tasks demonstrate the superiority of PRIME over the state-of-the-art models, especially with limited labels.

Dynamic graph learning equips the edges with time attributes and allows multiple links between two nodes, which is a crucial technology for understanding evolving data scenarios like traffic prediction and recommendation systems. Existing works obtain the evolving patterns mainly depending on the most recent neighbor sequences. However, we argue that whether two nodes will have interaction with each other in the future is highly correlated with the same interaction that happened in the past. Only considering the recent neighbors overlooks the phenomenon of repeat behavior and fails to accurately capture the temporal evolution of interactions. To fill this gap, this paper presents RepeatMixer, which considers evolving patterns of first and high-order repeat behavior in the neighbor sampling strategy and temporal information learning. Firstly, we define the first-order repeat-aware nodes of the source node as the destination nodes that have interacted historically and extend this concept to high orders as nodes in the destination node's high-order neighbors. Then, we extract neighbors of the source node that interacted before the appearance of repeat-aware nodes with a slide window strategy as its neighbor sequence. Next, we leverage both the first and high-order neighbor sequences of source and destination nodes to learn temporal patterns of interactions via an MLP-based encoder. Furthermore, considering the varying temporal patterns on different orders, we introduce a time-aware aggregation mechanism that adaptively aggregates the temporal representations from different orders based on the significance of their interaction time sequences. Experimental results demonstrate the superiority of RepeatMixer over state-of-the-art models in link prediction tasks, underscoring the effectiveness of the proposed repeat-aware neighbor sampling strategy.

Knowledge Tracing aims to assess student learning states by predicting their performance in answering questions. Different from the existing research which utilizes fixed-length learning sequence to obtain the student states and regards KT as a static problem, this work is motivated by three dynamical characteristics: 1) The scales of students answering records are constantly growing; 2) The semantics of time intervals between the records vary; 3) The relationships between students, questions and concepts are evolving. The three dynamical characteristics above contain the great potential to revolutionize the existing knowledge tracing methods. Along this line, we propose a Dynamic Graph-based Knowledge Tracing model, namely DyGKT. In particular, a continuous-time dynamic question-answering graph for knowledge tracing is constructed to deal with the infinitely growing answering behaviors, and it is worth mentioning that it is the first time dynamic graph learning technology is used in this field. Then, a dual time encoder is proposed to capture long-term and short-term semantics among the different time intervals. Finally, a multiset indicator is utilized to model the evolving relationships between students, questions, and concepts via the graph structural feature. Numerous experiments are conducted on five real-world datasets, and the results demonstrate the superiority of our model. All the used resources are publicly available at https://github.com/PengLinzhi/DyGKT.

Structure encoding has proven to be the key feature to distinguishing links in a graph. However, Structure encoding in the temporal graph keeps changing as the graph evolves, repeatedly computing such features can be time-consuming due to the high-order subgraph construction. We develop the Co-Neighbor Encoding Schema (CNES) to address this issue. Instead of recomputing the feature by the link, CNES stores information in the memory to avoid redundant calculations. Besides, unlike the existing memory-based dynamic graph learning method that stores node hidden states, we introduce a hashtable-based memory to compress the adjacency matrix for efficient structure feature construction and updating with vector computation in parallel. Furthermore, CNES introduces a Temporal-Diverse Memory to generate long-term and short-term structure encoding for neighbors with different structural information. A dynamic graph learning framework, Co-Neighbor Encoding Network (CNE-N), is proposed using the aforementioned techniques. Extensive experiments on thirteen public datasets verify the effectiveness and efficiency of the proposed method.

Patent classification aims to assign multiple International Patent Classification (IPC) codes to a given patent. Existing methods for automated patent classification primarily focus on analyzing the text descriptions of patents. However, apart from the textual information, each patent is also associated with some assignees, and the knowledge of their previously applied patents can often be valuable for accurate classification. Furthermore, the hierarchical taxonomy defined by the IPC system provides crucial contextual information and enables models to leverage the correlations between IPC codes for improved classification accuracy. However, existing methods fail to incorporate the above aspects and lead to reduced performance. To address these limitations, we propose an integrated framework that comprehensively considers patent-related information for patent classification. To be specific, we first present an IPC codes correlations learning module to capture both horizontal and vertical information within the IPC codes. This module effectively captures the correlations by adaptively exchanging and aggregating messages among IPC codes at the same level (horizontal information) and from both parent and children codes (vertical information), which allows for a comprehensive integration of knowledge and relationships within the IPC hierarchical taxonomy. Additionally, we design a historical application patterns learning component to incorporate previous patents of the corresponding assignee by aggregating high-order temporal information via a dual-channel graph neural network. Finally, our approach combines the contextual information from patent texts, which encompasses the semantics of IPC codes, with assignees’ sequential preferences to make predictions. Experimental evaluations on real-world datasets demonstrate the superiority of our proposed approach over existing methods. Moreover, we present the model’s ability to capture the temporal patterns of assignees and the semantic dependencies among IPC codes.

Forecasting traffic flow is an important task in urban areas, and a large number of methods have been proposed for traffic flow prediction. However, most of the existing methods follow a general technical route to aggregate historical information spatially and temporally. In this paper, we propose a different approach for traffic flow prediction. Our major motivation is to more effectively incorporate various intrinsic patterns in real-world traffic flows, such as fixed spatial distributions, topological correlations, and temporal periodicity. Along this line, we propose a novel autoencoder-based traffic flow prediction method, named Spatio-Temporal AutoEncoder (ST-AE). The core of our method is an autoencoder specially designed to learn the intrinsic patterns from traffic flow data, and encode the current traffic flow information into a low-dimensional representation. The prediction is made by simply projecting the current hidden states to the future hidden states, and then reconstructing the future traffic flows with the trained autoencoder. We have conducted extensive experiments on four real-world data sets. Our method outperforms existing methods in several settings, particularly for long-term traffic flow prediction.

Multivariate time series anomaly detection has great potentials in many practical applications. Extreme unbalanced training set and noise interference make it challenging to accurately capture the distribution of normal data and then detect anomalies. Existing AutoEncoder(AE)-based approaches are lack of effective regularization method specially designed for anomaly detection tasks thus easily overfitting while Generative Adversarial Network(GAN)-based approaches are mostly trained under the hypothesis of pollution-free training set, which means the training set is all composed of normal samples and that is hard to satisfy in practice. To tackle these problems, in this paper we propose a GAN based anomaly detection method for multivariate time series named FGANomaly (letter F is for Filter). The core idea is to filter possible anomalous samples with pseudo-labels before training the discriminator thus to capture the distribution of normal data as precise as possible. In addition, we design a novel training objective for the generator, which leads the generator to concentrate more on plausible normal data and ignore anomalies. We conducted comprehensive experiments on four public datasets, and the experimental results show the superiority of our method over baselines in both performance and robustness.

Recent studies have shown great promise in applying graph neural networks for multivariate time series forecasting, where the interactions of time series are described as a graph structure and the variables are represented as the graph nodes. Along this line, existing methods usually assume that the graph structure (or the adjacency matrix), which determines the aggregation manner of graph neural network, is fixed either by definition or self-learning. However, the interactions of variables can be dynamic and evolutionary in real-world scenarios. Furthermore, the interactions of time series are quite different if they are observed at different time scales. To equip the graph neural network with a flexible and practical graph structure, in this paper, we investigate how to model the evolutionary and multi-scale interactions of time series. In particular, we first provide a hierarchical graph structure cooperated with the dilated convolution to capture the scale-specific correlations among time series. Then, a series of adjacency matrices are constructed under a recurrent manner to represent the evolving correlations at each layer. Moreover, a unified neural network is provided to integrate the components above to get the final prediction. In this way, we can capture the pair-wise correlations and temporal dependency simultaneously. Finally, experiments on both single-step and multi-step forecasting tasks demonstrate the superiority of our method over the state-of-the-art approaches.

Graph Convolutional Network (GCN) has been widely applied in transportation demand prediction due to its excellent ability to capture non-Euclidean spatial dependence among station-level or regional transportation demands. However, in most of the existing research, the graph convolution was implemented on a heuristically generated adjacency matrix, which could neither reflect the real spatial relationships of stations accurately, nor capture the multi-level spatial dependence of demands adaptively. To cope with the above problems, this paper provides a novel graph convolutional network for transportation demand prediction. Firstly, a novel graph convolution architecture is proposed, which has different adjacency matrices in different layers and all the adjacency matrices are self-learned during the training process. Secondly, a layer-wise coupling mechanism is provided, which associates the upper-level adjacency matrix with the lower-level one. It also reduces the scale of parameters in our model. Lastly, a unitary network is constructed to give the final prediction result by integrating the hidden spatial states with gated recurrent unit, which could capture the multi-level spatial dependence and temporal dynamics simultaneously. Experiments have been conducted on two real-world datasets, NYC Citi Bike and NYC Taxi, and the results demonstrate the superiority of our model over the state-of-the-art ones.

Taxi and sharing bike bring great convenience to urban transportation. A lot of efforts have been made to improve the efficiency of taxi service or bike sharing system by predicting the next-period pick-up or drop-off demand. Different from the existing research, this paper is motivated by the following two facts: 1) From a micro view, an observed spatial demand at any time slot could be decomposed as a combination of many hidden spatial demand bases; 2) From a macro view, the multiple transportation demands are strongly correlated with each other, both spatially and temporally. Definitely, the above two views have great potential to revolutionize the existing taxi or bike demand prediction methods. Along this line, this paper provides a novel Co-prediction method based on Spatio-Temporal neural Network, namely, CoST-Net. In particular, a deep convolutional neural network is constructed to decompose a spatial demand into a combination of hidden spatial demand bases. The combination weight vector is used as a representation of the decomposed spatial demand. Then, a heterogeneous Long Short-Term Memory (LSTM) is proposed to integrate the states of multiple transportation demands, and also model the dynamics of them mixedly. Last, the environmental features such as humidity and temperature are incorporated with the achieved overall hidden states to predict the multiple demands simultaneously. Experiments have been conducted on real-world taxi and sharing bike demand data, results demonstrate the superiority of the proposed method over both classical and the state-of-the-art transportation demand prediction methods.