Дослідження методів побудови рекомендаційних систем з використанням графових нейронних мереж

Abstract

This work is devoted to the study of the application of Graph Neural Networks (GNNs) in the recommendation tasks. It summarizes the basic idea of GNN and its three main steps and outlines the relevance of research on recommender systems. It describes the LightGCN model, and its problem associated with the update method - using the aggregated representation of neighbors as the new central node representation, might overlook the intrinsic user preference or the intrinsic item property. Examples of variants of the solution of this issue are given.