ENHANCING GRAPH CONVOLUTIONAL NETWORK-BASED RECOMMENDATION SYSTEMS WITH CONTRASTIVE LOSS: DEVELOPMENT AND EXPERIMENTAL EVALUATION OF THE ROGER MODEL

This thesis presents the development and experimental evaluation of RoGER, a novel recommendation framework that integrates GCNs with contrastive learning to enhance robustness and personalization. The findings contribute to advancing the integration of graph neural networks and contrastive learning in modern recommender systems, with implications for more robust and explainable personalization.