Aug 2026· Journal of King Saud University: Computer and Information Sciences· Vol 38· 0 citations· 44 references
TL;DR
HGAT-Rec is proposed, which incorporates a heterogeneity-aware contrastive learning (HCL) objective that grounds view construction and sample selection in the typed relational structure of a cross-domain heterogeneous graph: type-stratified edge dropout preserves high-signal interaction channels proportionally to their attention weight.
Abstract
Cross-domain recommendation over heterogeneous e-commerce networks faces three unresolved technical failures: existing GNN recommenders apply a single shared attention vector across all neighbor pairs, which can introduce a type-conflation bias governed by the inter-type variance of attention coefficients; meta-path importance weights learned in a source domain become systematically miscalibrated under the relational distribution shift of cross-domain transfer; and contrastive learning methods apply type-agnostic perturbations that assign equal dropout intensity to semantically distinct edge types, degrading representation quality in heterogeneous graph settings. We propose HGAT-Rec, which incorporates a heterogeneity-aware contrastive learning (HCL) objective that grounds view construction and sample selection in the typed relational structure of a cross-domain heterogeneous graph: type-stratified edge dropout preserves high-signal interaction channels proportionally to their attention weight, type-conditioned positive samples are drawn from meta-path neighborhoods sharing genuine relational content, and a cross-domain InfoNCE term enforces relative ordering constraints across the full overlapping user population. To support HCL, HGAT-Rec further provides (i) a type-triple-indexed graph attention mechanism parameterized by the joint combination of source node type, target node type, and edge type; (ii) a domain-conditioned meta-path aggregation layer with separate per-domain path importance distributions sharing a domain-invariant query vector; and (iii) a TransR-based knowledge gated fusion module decoupling item semantics from interaction sparsity. Theoretical grounding is provided by two formal propositions characterizing type-conflation bias and contrastive robustness under sparsity. Experiments on the Amazon Product Co-purchasing benchmark show that HGAT-Rec outperforms eleven baselines, achieving 11.20% and 8.70% improvements in NDCG@10 and HR@10 over the strongest competitor, a 31.80% gain for cold-start users, and a sparsity degradation rate of 35.2% under 80% interaction removal versus 49.0% for the next-best baseline.
Personalized recommendation has become an essential component of intelligent information systems and electronic multimedia platforms. However, cold-start items with limited user–item interactions remain difficult to model, especially when collaborative signals are sparse and heterogeneous side information is underutilized. To address this problem, this paper proposes Heterogeneous Graph Contrastive Recommendation (HGCRec), a heterogeneous graph contrastive learning framework for cold-start multi-modal recommendation. HGCRec constructs a heterogeneous structural graph containing users, items, categories, brands, and semantic entities, while visual and textual information is incorporated as item-side feature views. A relation-aware heterogeneous graph encoder captures relation-specific structural semantics, and a cross-modal graph contrastive learning objective coordinates structural, visual, and textual item representations. An interaction-sparsity-aware weighting strategy allocates stronger contrastive supervision to items with fewer training interactions. Furthermore, an item-specific adaptive fusion module integrates collaborative, structural, visual, and textual representations according to interaction sparsity and learned multi-source representation states. Experiments are conducted on three Amazon multi-modal recommendation datasets, including Baby, Sports, and Clothing. The results show that HGCRec consistently outperforms representative graph-based and multi-modal recommendation baselines. Compared with the strongest baseline, Freezing and Denoising Graph Structures for Multimodal Recommendation (FREEDOM), HGCRec improves Recall@20 by 9.65%, 10.44%, and 12.00% on Baby, Sports, and Clothing, respectively, and improves NDCG@20 by 11.14%, 11.35%, and 13.10%. Sparsity-aware analysis further shows larger relative improvements for items with extremely limited training interactions, demonstrating the effectiveness of HGCRec under the evaluated interaction-sparse and few-shot settings.
A Cardinality-Decomposed Loss (CDL) is proposed that combines both Cross Entropy (CE) and BPR to enable the model to collectively optimize for relations across cardinalities and is found that CDL consistently improves discriminability in attribute embeddings.
Parul Maheshwari, Amulya Paruchuri, Yiqing Zou et al.· arXiv.org· 0 citations
Cross-domain e-commerce recommendation faces challenges from multimodal product heterogeneity, sparse intercategory associations, and opaque recommendation reasoning. To improve accuracy and interpretability, this study proposes a multimodal large-language-model-driven framework for self-evolving cross-domain product knowledge graphs and explainable recommendation. Product images, titles, and attributes from apparel, home-furnishing, and digital-product domains are encoded using a multimodal large model, mapped into a shared latent space, and aligned through contrastive learning for cross-domain entity and semantic association extraction. A self-evolution mechanism uses the large language model as a relation verifier and reasoning engine to validate, buffer, prune, or extend graph edges according to confidence and interaction feedback. A path-aware graph neural network then samples multi-hop cross-domain paths, encodes product sequences through gated recurrent units, and fuses graph representations with original multimodal embeddings. Recommendation explanations are generated from high-weight inference chains under loyalty constraints to ensure factual consistency. Experiments on a large e-commerce dataset show that the proposed method achieves Hits@10 of 0.892 and MRR of 0.537 for relation completion, while CTR, CVR, and Recall@20 reach 12.4%, 6.2%, and 22.1%, respectively. Explanation fidelity and perceived usefulness score 4.23 and 4.15. The framework supports semantic alignment, graph-based reasoning, and interpretable recommendation in multimodal information systems.
Li Ma, Yuanli Cui· Advanced Electromagnetics· 0 citations
This work proposes a self-expressive solver that captures the complementary homophily between meta-paths and node features to obtain ho-mophilous representations and designs separate path encoders to model diverse interactions, thus explicitly including cross-type interactions while mitigating noise via adaptive fusion.
Minda Chen, Yujie Mo, Jun-Kai Huang et al.· 0 citations
Graph Neural Networks (GNNs) and Large Language Models (LLMs) have each advanced recommendation systems by modeling structural and semantic signals, respectively. However, integrating their complementary strengths remains challenging, particularly in sparse settings where maintaining semantic precision is critical. We propose TRWH (Text-driven Random Walk Heterogeneous Graph Neural Network), a novel framework that fuses LLM-generated textual profiles with heterogeneous graph structures through strategic random walk augmentation. TRWH consists of three core components: (1) Embedding Creation, which produces user and item representations using both Word2Vec and LLM-based profiling; (2) a Heterogeneous Graph Neural Network (HeteroGNN) that propagates information across multi-relational edges; and (3) Random Walk-based Path Construction, which enriches sparse graphs with second-order user-user and item-item links. Experiments on the Amazon-2023 Fashion (2M users, 825K items) and Beauty (631K users, 112K items) datasets demonstrate that TRWH achieves substantial performance gains over state-of-the-art methods, including 80.0% RMSE and 52.6% MAE reductions on Fashion, and 25.7% and 10.8% improvements on Beauty. Notably, while random walks improve performance with traditional embeddings, they can dilute the nuanced representations learned by LLMs, underscoring the importance of adaptive integration strategies.
Multimodal Graph Neural Networks have become standard for recommendation by augmenting sparse interaction data with content features. Yet current architectures face two bottlenecks: structural rigidity, from a reliance on static precomputed similarity graphs that cannot adapt to evolving preferences; and semantic fragility, where noisy modality signals are indiscriminately fused, distorting the collaborative signal. We propose MURAL (Multimodal Uncertainty-aware Recommendation via Adaptive edge Learning), a unified framework that shifts multimodal recommendation from fixed structural augmentation to dynamic topology discovery. To address structural rigidity, an Adaptive Edge Learner combines a differentiable retrieval-augmented strategy with an approximate nearest neighbor search to discover latent item-item correlations that are both semantically adaptive and computationally scalable (O(NlogN)). To address semantic fragility, an Uncertainty-Aware Fusion module models the aleatoric uncertainty of heterogeneous modalities, dynamically down-weighting unreliable features while prioritizing high-confidence signals as a defense against cross-modal noise. We further employ a contrastive teacher-student alignment that anchors modality-specific representations to stable behavioral signals, ensuring optimization stability without gradient leakage. Experiments on large-scale benchmarks including TikTok and Amazon show that MURAL significantly surpasses both structural and generative state-of-the-art baselines, achieving superior accuracy while offering interpretability through domain-specific modality dominance and robustness under extreme data corruption.
Ahmad Mousavi (Department of Mathematics, Statistics American University), Majid Alikhani (Independent Researcher) et al.· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
