Abstract:Group recommendation has garnered significant attention recently, aiming to suggest items of interest to groups. Most deep learning-based approaches for group recommendation focus on learning group representation through group–user interactions or by aggregating user preferences. However, these approaches face challenges due to the complexity of group representation learning and the limited availability of group–item interactions. To address these difficulties, we propose a multi-graph neural group recommendation model with meta-learning and multi-teacher distillation. Our model consists of three stages: multiple graphs representation learning (MGRL), meta-learning-based knowledge transfer (MLKT), and multi-teacher distillation (MTD). In the MGRL stage, we construct two bipartite graphs and an undirected weighted graph based on group–user interactions, group–item interactions, and similarity between groups, respectively. We then utilize a linear variational graph auto-encoder network to learn the semantic features of groups, users, and items. In the MLKT stage, we introduce a multilayer perceptron (MLP) network for the recommendation task and design a meta-learning-based knowledge transfer algorithm to train an initialization parameter that incorporates user preference knowledge. In the MTD stage, we use the trained parameters as initialization parameters for all teacher networks. We evenly divide the group–item interactions among the teachers and sequentially train the teacher networks. Finally, we employ a multi-teacher distillation approach to train student networks with superior performance. Furthermore, we conduct extensive experiments on six real-world datasets to evaluate the effectiveness of our proposed model. The experimental findings demonstrate that our model, in comparison to state-of-the-art approaches, achieves superior performance in terms of group recommendation.

MetaGA: Metalearning with Graph-Attention for Improved Long-Tail Item Recommendation

Meta Concept Recommendation Based on Knowledge Graph

CoMeta: Enhancing Meta Embeddings with Collaborative Information in Cold-start Problem of Recommendation.

GACRec: Generative Adversarial Contrastive Learning for Improved Long-Tail Item Recommendation

RGMeta: Enhancing Cold-Start Recommendations with a Residual Graph Meta-Embedding Model

Long-tail Augmented Graph Contrastive Learning for Recommendation

MetaKRec: Collaborative Meta-Knowledge Enhanced Recommender System

A novel meta-graph-based attention model for event recommendation

Parameters Efficient Fine-Tuning for Long-Tailed Sequential Recommendation.

Long Short-Term Temporal Meta-learning in Online Recommendation

A General Tail Item Representation Enhancement Framework for Sequential Recommendation

MetaKG: Meta-learning on Knowledge Graph for Cold-start Recommendation

Combining Meta-Graph and Attention for Recommendation over Heterogenous Information Network

An improved sequential recommendation model based on spatial self-attention mechanism and meta learning

ICMT: Item Cluster-Wise Multi-Objective Training for Long-Tail Recommendation.

Multi-Behavior Enhanced Heterogeneous Graph Convolutional Networks Recommendation Algorithm based on Feature-Interaction

A Multi-Graph Neural Group Recommendation Model with Meta-Learning and Multi-Teacher Distillation

Deep Meta-learning in Recommendation Systems: A Survey

Graph Meta Network for Multi-Behavior Recommendation

A Reinforcement Learning List Recommendation Model Fused with Graph Neural Networks

MMGCL: Meta Knowledge-Enhanced Multi-view Graph Contrastive Learning for Recommendations