A well-informed recommendation framework could not only help users identify their interested items, but also benefit the revenue of various online platforms (e.g., e-commerce, social media). Traditional models usually assume that a single type interaction exists between user and item, fail to model multiplex user-item relationships from multi-typed behavior data, such as page view, add-to-favourite purchase. While some recent studies propose capture dependencies across different types...

Graph Neural Networks (GNNs) have become powerful tools in modeling graph-structured data recommender systems. However, real-life recommendation scenarios usually involve heterogeneous relationships (e.g., social-aware user influence, knowledge-aware item dependency) which contains fruitful information to enhance the preference learning. In this paper, we study problem of graph-enhanced relational learning for recommendation. Recently, contrastive self-supervised has successful light this,...

The online emergence of multi-modal sharing platforms (eg, TikTok, Youtube) is powering personalized recommender systems to incorporate various modalities visual, textual and acoustic) into the latent user representations. While existing works on recommendation exploit multimedia content features in enhancing item embeddings, their model representation capability limited by heavy label reliance weak robustness sparse behavior data. Inspired recent progress self-supervised learning...

The problem of data sparsity has long been a challenge in recommendation systems, and previous studies have attempted to address this issue by incorporating side information. However, approach often introduces effects such as noise, availability issues, low quality, which turn hinder the accurate modeling user preferences adversely impact performance. In light recent advancements large language models (LLMs), possess extensive knowledge bases strong reasoning capabilities, we propose novel...

Recommender systems have seen significant advancements with the influence of deep learning and graph neural networks, particularly in capturing complex user-item relationships.However, these graph-based recommenders heavily depend on ID-based data, potentially disregarding valuable textual information associated users items, resulting less informative learned representations.Moreover, utilization implicit feedback data introduces potential noise bias, posing challenges for effectiveness user...

Self-supervised learning (SSL) has gained significant interest in recent years as a solution to address the challenges posed by sparse and noisy data recommender systems. Despite growing number of SSL algorithms designed provide state-of-the-art performance various recommendation scenarios (e.g., graph collaborative filtering, sequential recommendation, social KG-enhanced recommendation), there is still lack unified frameworks that integrate across different domains. Such framework could...

Sequential recommendation (SR) aims to predict the subsequent behaviors of users by understanding their successive historical behaviors. Recently, some methods for SR are devoted alleviating data sparsity problem (i.e., limited supervised signals training), which take account contrastive learning incorporate self-supervised into SR. Despite achievements, it is far from enough learn informative user/item embeddings due inadequacy modeling complex collaborative information and co-action...

Graph Structure Learning (GSL) focuses on capturing intrinsic dependencies and interactions among nodes in graph-structured data by generating novel graph structures. Neural Networks (GNNs) have emerged as promising GSL solutions, utilizing recursive message passing to encode node-wise inter-dependencies. However, many existing methods heavily depend explicit structural information supervision signals, leaving them susceptible challenges such noise sparsity. In this work, we propose...

Recommender systems play a crucial role in tackling the challenge of information overload by delivering personalized recommendations based on individual user preferences. Deep learning techniques, such as RNNs, GNNs, and Transformer architectures, have significantly propelled advancement recommender enhancing their comprehension behaviors However, supervised methods encounter challenges real-life scenarios due to data sparsity, resulting limitations ability learn representations effectively....

Heterogeneous graph learning aims to capture complex relationships and diverse relational semantics among entities in a heterogeneous obtain meaningful representations for nodes edges. Recent advancements neural networks (HGNNs) have achieved state-of-the-art performance by considering relation heterogeneity using specialized message functions aggregation rules. However, existing frameworks limitations generalizing across datasets. Most of these follow the "pre-train" "fine-tune" paradigm on...

Personalized recommender systems play a crucial role in capturing users' evolving preferences over time to provide accurate and effective recommendations on various online platforms. However, many recommendation models rely single type of behavior learning, which limits their ability represent the complex relationships between users items real-life scenarios. In such situations, interact with multiple ways, including clicking, tagging as favorite, reviewing, purchasing. To address this...

The problem of data sparsity has long been a challenge in recommendation systems, and previous studies have attempted to address this issue by incorporating side information. However, approach often introduces effects such as noise, availability issues, low quality, which turn hinder the accurate modeling user preferences adversely impact performance. In light recent advancements large language models (LLMs), possess extensive knowledge bases strong reasoning capabilities, we propose novel...

Recommender systems have seen significant advancements with the influence of deep learning and graph neural networks, particularly in capturing complex user-item relationships. However, these graph-based recommenders heavily depend on ID-based data, potentially disregarding valuable textual information associated users items, resulting less informative learned representations. Moreover, utilization implicit feedback data introduces potential noise bias, posing challenges for effectiveness...

Heterogeneous graph learning aims to capture complex relationships and diverse relational semantics among entities in a heterogeneous obtain meaningful representations for nodes edges. Recent advancements neural networks (HGNNs) have achieved state-of-the-art performance by considering relation heterogeneity using specialized message functions aggregation rules. However, existing frameworks limitations generalizing across datasets. Most of these follow the "pre-train" "fine-tune" paradigm on...

Despite the remarkable ability of large vision-language models (LVLMs) in image comprehension, these frequently generate plausible yet factually incorrect responses, a phenomenon known as hallucination.Recently, language (LLMs), augmenting LLMs by retrieving information from external knowledge resources has been proven promising solution to mitigate hallucinations.However, retrieval augmentation LVLM significantly lags behind widespread applications LVLM. Moreover, when transferred LVLMs,...

Spatio-temporal prediction is a crucial research area in data-driven urban computing, with implications for transportation, public safety, and environmental monitoring. However, scalability generalization challenges remain significant obstacles. Advanced models often rely on Graph Neural Networks to encode spatial temporal correlations, but struggle the increased complexity of large-scale datasets. The recursive GNN-based message passing schemes used these hinder their training deployment...

Spatio-temporal prediction is a crucial research area in data-driven urban computing, with implications for transportation, public safety, and environmental monitoring. However, scalability generalization challenges remain significant obstacles. Advanced models often rely on Graph Neural Networks to encode spatial temporal correlations, but struggle the increased complexity of large-scale datasets. The recursive GNN-based message passing schemes used these hinder their training deployment...

Recently, generative pre-training based models have demonstrated remarkable results on Aspect-based Sentiment Analysis (ABSA) task. However, previous works overemphasize crafting various templates to paraphrase training targets for enhanced decoding, ignoring the internal optimizations models. Despite notable achieved by these target-oriented optimization methods, they struggle with complicated long texts since implicit long-distance relation, e.g., aspect-opinion is difficult extract under...

Self-supervised learning (SSL) has gained significant interest in recent years as a solution to address the challenges posed by sparse and noisy data recommender systems. Despite growing number of SSL algorithms designed provide state-of-the-art performance various recommendation scenarios (e.g., graph collaborative filtering, sequential recommendation, social KG-enhanced recommendation), there is still lack unified frameworks that integrate across different domains. Such framework could...