Graph Neural Networks for RecSys @ LoG 2023

[UPDATE] We kindly ask you to fill in this anonymous form regarding the overall quality of the tutorial. As we are planning to present it to other venues, your feedback might be extremely important for us 🙏🙂

Abstract

Graph neural networks (GNNs) have gained prominence in recommendation systems in recent years. By representing the user-item matrix as a bipartite and undirected graph, GNNs have demonstrated their potential to capture short- and long-distance user-item interactions, thereby learning more accurate preference patterns than traditional recommendation approaches. In contrast to previous tutorials on the same topic, this tutorial aims to present and examine three key aspects that characterize GNNs for recommendation: (i) the reproducibility of state-of-the-art approaches, (ii) the potential impact of graph topological characteristics on the performance of these models, and (iii) strategies for learning node representations when training features from scratch or utilizing pre-trained embeddings as additional item information (e.g., multimodal features). The goal is to provide three novel theoretical and practical perspectives on the field, currently subject to debate in graph learning but long been overlooked in the context of recommendation systems.

Tutorial schedule

Total tutorial duration: 180 minutes

Pt.0 Introduction Pt.1 Reproducibility Pt.2 Graph Topology Pt.3 Node Representation

Introduction and background (Tommaso Di Noia): 20 minutes

Introduction and motivations of the tutorial: 5 minutes
Basics concepts of recommender systems & GNNs-based recommendation: 15 minutes

Reproducibility (Claudio Pomo): 60 minutes

[Hands-on #1] Implementation and reproducibility of GNNs-based recsys in Elliot with PyG and reproducibility issues: 35 minutes
Performance comparison of GNNs-based approaches to traditional recommendation systems: 25 minutes

Break and Q&A: 15 minutes

Graph topology: 30 minutes

Concepts and formulations of graph topological properties of the user-item graph (Tommaso Di Noia): 15 minutes
Impact of topological graph properties on the performance of GNNs-based recommender systems (Daniele Malitesta): 15 minutes

Node representation (Daniele Malitesta): 45 minutes

Design choices to train node embeddings from scratch: 20 minutes
[Hands-on #2] Leveraging item’s side-information (e.g., multimodal features) to represent node embeddings: 25 minutes

Closing remarks and Q&A: 10 minutes

Additional useful material

Title	Paper	Slides	Code	Venue	Year
How Neighborhood Exploration Influences Novelty and Diversity in Graph Collaborative Filtering	link	link	link	MORS @ RecSys	2022
Auditing Consumer- and Producer-Fairness in Graph Collaborative Filtering	link	link	link	ECIR	2023
An Out-of-the-Box Application for Reproducible Graph Collaborative Filtering extending the Elliot Framework	link	link	link	UMAP/GLB @ KDD	2023
Challenging the Myth of Graph Collaborative Filtering: a Reasoned and Reproducibility-driven Analysis	link	link	link	RecSys	2023
A Topology-aware Analysis of Graph Collaborative Filtering	link		link	arXiv	2023
Disentangling the Performance Puzzle of Multimodal-aware Recommender Systems	link	link	link	EvalRS@KDD	2023
On Popularity Bias of Multimodal-aware Recommender Systems: a Modalities-driven Analysis	link	link	link	MMIR @ MM	2023
Formalizing Multimedia Recommendation through Multimodal Deep Learning	link		link	arXiv	2023