Search

Dynamic recommendation systems aim to achieve real-time updates and dynamic migration of user interests, primarily utilizing user-item interaction sequences with timestamps to capture the dynamic changes in user interests and item attributes. Recent research has mainly centered on two aspects. First, it involves modeling the dynamic interaction relationships between users and items using dynamic graphs. Second, it focuses on mining their long-term and short-term interaction patterns. This is achieved through the joint learning of static and dynamic embeddings for both users and items. Although most existing methods have achieved some success in modeling the historical interaction sequences between users and items, there is still room for improvement, particularly in terms of modeling the long-term dependency structures of dynamic interaction histories and extracting the most relevant delayed interaction patterns. To address this issue, we proposed a Dynamic Context-Aware Recommendation System for dynamic recommendation. Specifically, our model is built on a dynamic graph and utilizes the static embeddings of recent user-item interactions as dynamic context. Additionally, we constructed a Gated Multi-Layer Perceptron encoder to capture the long-term dependency structure in the dynamic interaction history and extract high-level features. Then, we introduced an Attention Pooling network to learn similarity scores between high-level features in the user-item dynamic interaction history. By calculating bidirectional attention weights, we extracted the most relevant delayed interaction patterns from the historical sequence to predict the dynamic embeddings of users and items. Additionally, we proposed a loss function called the Pairwise Cosine Similarity loss for dynamic recommendation to jointly optimize the static and dynamic embeddings of two types of nodes. Finally, extensive experiments on two real-world datasets, LastFM, and the Global Terrorism Database showed that our model achieves consistent improvements over state-of-the-art baselines.

Abstract Graph neural networks (GNNs) have significant advantages in dealing with non-Euclidean data and have been widely used in various fields. However, most of the existing GNN models face two main challenges: (1) Most GNN models built upon the message-passing framework exhibit a shallow structure, which hampers their ability to efficiently transmit information between distant nodes. To address this, we aim to propose a novel message-passing framework, enabling the construction of GNN models with deep architectures akin to convolutional neural networks (CNNs), potentially comprising dozens or even hundreds of layers. (2) Existing models often approach the learning of edge and node features as separate tasks. To overcome this limitation, we aspire to develop a deep graph convolutional neural network learning framework capable of simultaneously acquiring edge embeddings and node embeddings. By utilizing the learned multi-dimensional edge feature matrix, we construct multi-channel filters to more effectively capture accurate node features. To address these challenges, we propose the Co-embedding of Edges and Nodes with Deep Graph Convolutional Neural Networks (CEN-DGCNN). In our approach, we propose a novel message-passing framework that can fully integrate and utilize both node features and multi-dimensional edge features. Based on this framework, we develop a deep graph convolutional neural network model that prevents over-smoothing and obtains node non-local structural features and refined high-order node features by extracting long-distance dependencies between nodes and utilizing multi-dimensional edge features. Moreover, we propose a novel graph convolutional layer that can learn node embeddings and multi-dimensional edge embeddings simultaneously. The layer updates multi-dimensional edge embeddings across layers based on node features and an attention mechanism, which enables efficient utilization and fusion of both node and edge features. Additionally, we propose a multi-dimensional edge feature encoding method based on directed edges, and use the resulting multi-dimensional edge feature matrix to construct a multi-channel filter to filter the node information. Lastly, extensive experiments show that CEN-DGCNN outperforms a large number of graph neural network baseline methods, demonstrating the effectiveness of our proposed method.

Abstract Although numerous spatiotemporal approaches have been presented to address the problem of missing spatiotemporal data, there are still limitations in concurrently capturing the underlying spatiotemporal dependence of spatiotemporal graph data. Furthermore, most imputation methods miss the hidden dynamic connection associations that exist between graph nodes over time. To address the aforementioned spatiotemporal data imputation challenge, we present an attention-based message passing and dynamic graph convolution network (ADGCN). Specifically, this paper uses attention mechanisms to unify temporal and spatial continuity and aggregate node neighbor information in multiple directions. Furthermore, a dynamic graph convolution module is designed to capture constantly changing spatial correlations in sensors utilizing a new dynamic graph generation method with gating to transmit node information. Extensive imputation tests in the air quality and traffic flow domains were carried out on four real missing data sets. Experiments show that the ADGCN outperforms the state-of-the-art baseline.

The dynamic recommender system realizes the real-time recommendation for users by learning the dynamic interest characteristics, which is especially suitable for the scenarios of rapid transfer of user interests, such as e-commerce and social media. The dynamic recommendation model mainly depends on the user-item history interaction sequence with timestamp, which contains historical records that reflect changes in the true interests of users and the popularity of items. Previous methods usually model interaction sequences to learn the dynamic embedding of users and items. However, these methods can not directly capture the excitation effects of different historical information on the evolution process of both sides of the interaction, i.e., the ability of events to influence the occurrence of another event. In this work, we propose a Dynamic Graph Hawkes Process based on Linear complexity Self-Attention (DGHP-LISA) for dynamic recommender systems, which is a new framework for modeling the dynamic relationship between users and items at the same time. Specifically, DGHP-LISA is built on dynamic graph and uses Hawkes process to capture the excitation effects between events. In addition, we propose a new self-attention with linear complexity to model the time correlation of different historical events and the dynamic correlation between different update mechanisms, which drives more accurate modeling of the evolution process of both sides of the interaction. Extensive experiments on three real-world datasets show that our model achieves consistent improvements over state-of-the-art baselines.

Graph Convolutional Neural Networks (GCNs) have been widely used in various fields due to their powerful capabilities in processing graph-structured data. However, GCNs encounter significant challenges when applied to scale-free graphs with power-law distributions, resulting in substantial distortions. Moreover, most of the existing GCN models are shallow structures, which restricts their ability to capture dependencies among distant nodes and more refined high-order node features in scale-free graphs with hierarchical structures. To more broadly and precisely apply GCNs to real-world graphs exhibiting scale-free or hierarchical structures and utilize multi-level aggregation of GCNs for capturing high-level information in local representations, we propose the Hyperbolic Deep Graph Convolutional Neural Network (HDGCNN), an end-to-end deep graph representation learning framework that can map scale-free graphs from Euclidean space to hyperbolic space. In HDGCNN, we define the fundamental operations of deep graph convolutional neural networks in hyperbolic space. Additionally, we introduce a hyperbolic feature transformation method based on identity mapping and a dense connection scheme based on a novel non-local message passing framework. In addition, we present a neighborhood aggregation method that combines initial structural featureswith hyperbolic attention coefficients. Through the abovemethods, HDGCNN effectively leverages both the structural features and node features of graph data, enabling enhanced exploration of non-local structural features and more refined node features in scale-free or hierarchical graphs. Experimental results demonstrate that HDGCNN achieves remarkable performance improvements over state-ofthe-art GCNs in node classification and link prediction tasks, even when utilizing low-dimensional embedding representations. Furthermore, when compared to shallow hyperbolic graph convolutional neural network models, HDGCNN exhibits notable advantages and performance enhancements. [ABSTRACT FROM AUTHOR]