Statistical Properties of Graph Neural Networks
Project with Owen Ward
Graph neural networks (GNNs) are an extremely popular and powerful tool in machine learning. However, the statistical properties of these methods have not been thoroughly explored. In this project a student will carry out a review of GNN's and the existing literature. They will conduct an empirical study of graph neural networks, evaluating and documenting their performance across a variety of common machine learning tasks on network data. Strong programming experience (Python, R) is required and previous experience with a modern machine learning library such as torch is preferred.