Road network representation learning (RNRL) has attracted increasing attention from both researchers and practitioners as various spatiotemporal tasks are emerging. Recent advanced methods leverage Graph Neural Networks (GNNs) and contrastive learning to characterize the spatial structure of road segments in a self-supervised paradigm. However, spatial heterogeneity and temporal dynamics of road networks raise severe challenges to the neighborhood smoothing mechanism of self-supervised GNNs. To address these issues, we propose a Dual-branch Spatial-Temporal self-supervised representation framework for enhanced road representations, termed as DST. On one hand, DST designs a mix-hop transition matrix for graph convolution to incorporate dynamic relations of roads from trajectories. Besides, DST contrasts road representations of the vanilla road network against that of hypergraphs in a spatial self-supervised way. The hypergraph is newly built based on three types of hyperedges to capture long-range relations. On the other hand, DST performs next token prediction as the temporal self-supervised task on the sequences of traffic dynamics based on a causal Transformer, which is further regularized by differentiating traffic modes of weekdays from those of weekends. Extensive experiments against state-of-the-art methods verify the superiority of our proposed framework. Moreover, the comprehensive spatiotemporal modeling facilitates DST to excel in zero-shot learning scenarios.

Recent years have seen a rapid surge in research leveraging Large Language Models (LLMs) for recommendation. These methods typically employ supervised fine-tuning (SFT) to adapt LLMs to recommendation scenarios, and utilize beam search during inference to efficiently retrieve B top-ranked recommended items. However, we identify a critical training-inference inconsistency: while SFT optimizes the overall probability of positive items, it does not guarantee that such items will be retrieved by beam search even if they possess high overall probabilities. Due to the greedy pruning mechanism, beam search can prematurely discard a positive item once its prefix probability is insufficient.

Recommendation systems (RS) aim to retrieve the top-K items most relevant to users, with metrics such as Precision@K and Recall@K commonly used to assess effectiveness. The architecture of an RS model acts as an inductive bias, shaping the patterns the model is inclined to learn. In recent years, numerous recommendation architectures have emerged, spanning traditional matrix factorization, deep neural networks, and graph neural networks. However, their designs are often not explicitly aligned with the top-K objective, thereby limiting their effectiveness. To address this limitation, we propose TopKGAT, a novel recommendation architecture directly derived from a differentiable approximation of top-K metrics. The forward computation of a single TopKGAT layer is intrinsically aligned with the gradient ascent dynamics of the Precision@K metric, enabling the model to naturally improve top-K recommendation accuracy. Structurally, TopKGAT resembles a graph attention network and can be implemented efficiently. Extensive experiments on four benchmark datasets demonstrate that TopKGAT consistently outperforms state-of-the-art baselines. The code is available at https://github.com/StupidThree/TopKGAT.

Partially Relevant Video Retrieval (PRVR) is a challenging task in the domain of multimedia retrieval. It is designed to identify and retrieve untrimmed videos that are partially relevant to the provided query. In this work, we investigate long-sequence video content understanding to address information redundancy issues. Leveraging the outstanding long-term state space modeling capability and linear scalability of the Mamba module, we introduce a multi-Mamba module with temporal fusion framework (MamFusion) tailored for PRVR task. This framework effectively captures the state-relatedness in long-term video content and seamlessly integrates it into text-video relevance understanding, thereby enhancing the retrieval process. Specifically, we introduce Temporal T-to-V Fusion and Temporal V-to-T Fusion to explicitly model temporal relationships between text queries and video moments, improving contextual awareness and retrieval accuracy. Extensive experiments conducted on large-scale datasets demonstrate that MamFusion achieves state-of-the-art performance in retrieval effectiveness. Code is available at the link: https://github.com/Vision-Multimodal-Lab-HZCU/MamFusion.

Multimodal perception, which integrates vision and touch, is increasingly demonstrating its significance in domains such as embodied intelligence and human-computer interaction. However, in open-world scenarios, multimodal data streams face significant challenges, including catastrophic forgetting and overfitting, during few-shot class incremental learning (FSCIL), leading to a severe degradation in model performance. In this work, we propose a novel approach named Few-Shot Incremental Multi-modal Learning via Touch Guidance and Imaginary Vision Synthesis (TIFS). Our method leverages vision imagination synthesis to enhance the semantic understanding and integrates touch and vision fusion to improve the problem of modal imbalance. Specifically, we introduce a framework that employs touch-guided vision information for cross-modal contrastive learning to address the challenges of few-shot learning. Additionally, we incorporate multiple learning mechanisms, including regularization, memory mechanisms, and attention mechanisms, to mitigate catastrophic forgetting during multi-incremental step learning. Experimental results on the Touch and Go and VisGel datasets demonstrate that the TIFS framework exhibits robust continuous learning capabilities and strong generalization performance in touch-vision few-shot incremental learning tasks. Our code is available at https://github.com/Vision-Multimodal-Lab-HZCU/TIFS.

Current high-performance multi-modal models are predominantly trained in a static learning scenario, where the model undergoes a single joint training on the entire dataset. However, in real-world applications, data are dynamically generated, and tasks evolve continuously, necessitating that multi-modal models adapt to a continual learning scenario. The core challenge in continuous learning is catastrophic forgetting problem, where training data from previous tasks are often not fully retained. As a result, when the model learns new tasks, it can only utilize the training data associated with these new tasks, leading to a significant degradation in performance on previously learned tasks. Specifically, due to the costly process of collecting touch data and the low standardization of sensor outputs, tactile-visual continual learning poses significant challenges. To effectively integrate tactile and visual information and mitigate catastrophic forgetting, this paper proposes the Tactile-Visual Class-Continual Learning via Temporal Attention Interaction (TV-CCL) model. TV-CCL maintains the instance and class-level semantic similarity between tactile and visual modalities through Tactile-Visual Multi-modal Semantic Alignment (TV-MSA). Additionally, we incorporate Tactile-Guided Visual Attention Distillation (TG-VAD) to preserve previously learned haptic-guided visual attention capabilities. Our experiments on the Touch and Go dataset demonstrate that TV-CCL significantly outperforms existing CCL methods when leveraging combined haptic and visual information. Our code is available at https://github.com/Vision-Multimodal-Lab-HZCU/TV-CCL.

The popularity of mobile games has surged in recent years. Along with mobile games, the emergence of mini-games has recently raised attention. Compared to traditional mobile games, mini-games are more lightweight and platform-independent with low development cost, which has attracted thousands of developers and users. WeChat mini-games platform is one of the most popular platforms with over 100 000 mini-games. The diversity and variety of WeChat mini-games make it an ideal platform for training reinforcement learning (RL) agents. In contrast, most of the existing RL benchmark environments are equipped with predetermined games, which are always limited to several genres and lack the utilization of new and diverse games. To utilize the WeChat mini-games for RL research, in this article, we propose WemiEnv, a lightweight, easy-to-use and open-source platform for RL research towards WeChat mini-games. WemiEnv is built on the WeChat developer tools and allows RL agents to interact with the mini-games. WemiEnv also supports user-customized mini-games, requiring users to implement only a few interface functions within WemiEnv API. We also provide six popular mini-games:Space Fighter,Flip, 2048,Flappy Bird,Timberman, andSnakeas ready-to-use tasks. Experiments were conducted with the OpenAI Spinning Up library for RL baselines on the provided tasks to test the usability of WemiEnv.

Adversarial training (AT) is currently one of the most effective ways to obtain the robustness of deep neural networks against adversarial attacks. However, most AT methods suffer from robust overfitting, i.e., a significant generalization gap in adversarial robustness between the training and testing curves. In this paper, we first identify a connection between robust overfitting and the excessive memorization of noisy labels in AT from a view of gradient norm. As such label noise is mainly caused by a distribution mismatch and improper label assignments, we are motivated to propose a label refinement approach for AT. Specifically, our Self-Guided Label Refinement first self-refines a more accurate and informative label distribution from over-confident hard labels, and then it calibrates the training by dynamically incorporating knowledge from self-distilled models into the current model and thus requiring no external teachers. Empirical results demonstrate that our method can simultaneously boost the standard accuracy and robust performance across multiple benchmark datasets, attack types, and architectures. In addition, we also provide a set of analyses from the perspectives of information theory to dive into our method and suggest the importance of soft labels for robust generalization.

The next location recommendation is a significant task in spatio-temporal data mining (STDM), leading to an increased interest in the inherent dynamics in large-scale trajectory data. However, existing methods often prioritize the transition of locations while overlooking the collaborative signals between users, resulting in further impacts on modeling the higher-order effects between locations. Additionally, they struggle to extract useful patterns from long sequences, let alone capturing the long sequences that demonstrate the collaborative effects between users and locations. In light of these challenges, we construct a temporal graph based on the order of user movement, enabling both users and locations to utilize collaborative filtering signals. This greatly alleviates issues related to data sparsity and high-quality representation. Simultaneously, we introduce a Dual-View Graph Transformer model(TrajGraph), which samples sequences from spatial and temporal views by a dual-view sequence sampling method, independently encoding each view with a graph transformer to obtain effective representations of visited nodes, effectively addressing the high complexity of the transformer and ensuring efficiency and effectiveness. Extensive experiments on three public location-based service datasets demonstrate that our model can consistently outperform all baselines. Elaborate ablation studies further prove the effectiveness of spatial and temporal factors.

Block-based programming is of great significance for cultivating children’s computational thinking. However, due to the following challenges, it is difficult to evaluate students’ programming ability in online learning systems: 1) compared with the traditional Online Judge (OJ) system, there is no standard answer for a given task in block-based programming; 2) in order to promote students’ interests, although the programs are not totally correct and unrelated to the task, the teacher will give a comparatively higher score. Therefore, current approaches involving output comparison and code analysis do not work effectively. Furthermore, deep learning methods also suffer from the problem of how to represent block code for classification. We propose a novel hybrid estimation model to address these challenges. We first learn graph embedding from the parsed Abstract Syntax Tree (AST) to present the logicality of the code. Next, we provide some methods to measure the workload and complexity of the code. Then, we extracted some key variables and task-irrelevant properties, introduced teacher bias. Finally, XGBoost was constructed for classification. Based on real-world data collected from an online Scratch platform by early-age students, our model outperforms KimCNN, ResNet-18, and Graph2Vec+XGBoost. Moreover, we provided statistical analyses and intuitive explanations to interpret the characteristics in various groups.

Human trajectory prediction is an essential task for various applications such as travel recommendation, location-sensitive advertisement, and traffic planning. Most existing approaches are sequential-model based and produce a prediction by mining behavior patterns. However, the effectiveness of pattern-based methods is not as good as expected in real-life conditions, such as data sparse or data missing. Moreover, due to the technical limitations of sensors or the traffic situation at the given time, people going to the same place may produce different trajectories. Even for people traveling along the same route, the observed transit records are not exactly the same. Therefore trajectories are always diverse, and extracting user intention from trajectories is difficult. In this paper, we propose an augmented-intention recurrent neural network (AI-RNN) model to predict locations in diverse trajectories. We first propose three strategies to generate graph structures to demonstrate travel context and then leverage graph convolutional networks to augment user travel intentions under graph view. Finally, we use gated recurrent units with augmented node vectors to predict human trajectories. We experiment with two representative real-life datasets and evaluate the performance of the proposed model by comparing its results with those of other state-of-the-art models. The results demonstrate that the AI-RNN model outperforms other methods in terms of top-k accuracy, especially in scenarios with low similarity.

As adaptive Web applications are gaining importance in software domains nowadays, there is a need for an effective method to construct such applications. In this paper, we propose MAWA, a method for model-driven development of adaptive Web applications. MAWA considers the architecture of adaptive Web applications and the key elements of the development activity. It adopts an iterative, incremental development process, and the adaptive model, which is composed of context model and user model, is highlighted. Besides, the adaptive categories and mechanism are specified in MAWA, which are used to support the implementation of the adaptive behaviors. A code generation strategy is integrated into MAWA, and it can help us to produce the applications quickly and effectively.

An original GA (genetic algorithm) is usually used for QoS (quality of service)-based Web services selection, however, such algorithm has to do a large redundancy repeat to find a solution for the reason that it does not make use of adequate output information. Thus, the efficiency and precision of GA are reduced. Moreover, fixed fitness function can not change to fit for different composite situations and complete automatic process so GA sometimes would miss the most suitable solution. To remedy this situation, this paper proposes a novel algorithm GBAA (genetic based ant algorithm) which could put feedback information to original GA by using MMAS (max-min ant system) and it also overcome some drawbacks of MMAS such as long time need and slow convergence. The new approach could gain the benefits of both GA and MMAS. Besides GBAA DC (divide and composite) method and rank sort method are added. DC is adopted to analysis business requirement and to build process. After computing QoS attributes, a manual step is allowed to put into selection process and a rank sort method is used to distinguish profits of different solutions. The automatic process which is cooperated with manual process could help user to make more effective decisions on Web service selection.

Web services and SOA technologies are growing with a fast rate but still facing many problems due to their heterogeneous nature. This paper, based on OWL-S, presents a rich and extensible ontology framework named OWL-QSP for service compositions. In the framework, Service Type is imported to improve service abstract level, and QoS, situation, context are adopted. Since service discovery, service selection and service execution can adapt to the changing situation, QoS and situation-aware service-based systems are more dynamic and flexible so to better satisfy the users' functional and non-functional requirements. The introducing of policy permits managing WSs at a high level and facilitate reuse. It also presents SMICE, a prototype of the service composition system, and describes its main components with service composition process.