Embedding Theorey in Collaborative Filtering

Materials INFO

• Semester: Fall, 2025
• This page provides links to papers only.

CONTENTS

• [Week 1] Multiplex Graph Neural Network for Recommendation Task
  • Multi-relation graph contrastive learning with adaptive strategy for social recommendation (2025) (link)
    • Creates a noise-contrastive view for each relation to supplement sparse edge-splitting loss.
  • Relation-aware multiplex heterogeneous graph neural network (2025) (link)
    • Proposes a relation-aware encoder that fuses base edge matrices into hybrid relations and focuses on differences among edge types.
  • Multi-behavior recommendation with SVD Graph Neural Networks (2024) (link)
    • Builds SVD-based subgraphs to extract low-dimensional singular-value spaces for each behavior (view), then combines them with GNN message passing; effective in sparse multi-behavior scenarios.
  • Dual-Channel Multiplex Graph Neural Networks for Recommendation (2024) (link)
    • Separates explicit behavior-pattern and implicit relation-chain channels to model behavior order; improves mean Recall@10 by 10% on three real datasets.
  • Multi-Behavior Graph Neural Networks for Recommender System (2023) (link)
    • Constructs a multiplex graph with a layer for each behavioral edge type and learns interactions among relations using gated aggregation and a mutual-relation encoder.
  • Multiplex Heterogeneous Graph Neural Network with Behavior Pattern Modeling (2023) (link)
    • Separates multiple user–item behaviors, such as clicks and purchases, into pattern blocks, then aggregates interactions within patterns using behavior-aware attention to improve recommendation accuracy.
• [Week 2]
  • Graphon Estimation in bipartite graphs with observable edge labels and unobservable node labels (2023) (link)
    • Proposes a method for estimating the conditional expectation of a data matrix (a graphon) given latent variables. It analyzes the theoretical properties of an LSE estimator that cannot be implemented in practice, and demonstrates the empirical performance of a graphon estimator obtained by adapting Lloyd’s alternating minimization algorithm.