mutag

Pytorch implementation of Relational GCN for node classification

Explainable Graph Neural Networks (XGNNs) and Reinforcement Learning (RL) techniques for graph generation and optimization tasks. The project aims to explore and validate the use of explainable AI for creating interpretable graph structures.

Graph Convolution Neural Networks for Mutagenic Molecule Detection

Estimating graph vertex/edge counts from rendered images using a CNN, benchmarked against a GCN baseline and a density heuristic — with Grad-CAM, shortcut-learning probes, and failure-case analysis of what the model actually learns. Includes a deployed Streamlit demo app.

Comparative study of graph classification methods on MUTAG using graph kernels, embeddings, and GNNs.