Modeling and Algorithmic Advances for Random Dot Product Graphs - Bernardo Marenco (2025)

The Random Dot Product Graph (RDPG) model has emerged as a fundamental tool for representing network data through latent position embeddings. In this framework, each node is associated with a low-dimensional vector, and edges are formed with probabilities given by the inner products of these latent positions. This thesis investigates both the theoretical foundations and algorithmic aspects of inference under the RDPG model, with a focus on advancing graph representation learning for statistical network analysis.