This project is about learning latent graph structures for time-series datasets, which will then be used for forecasting and time-series anomaly detection tasks with graph neural networks. To achieve this goal, I need to run several experiments trying out existing methods and modifying them for better performance.