PyTorch Implementation of RouteNet

• RouteNet models a computer network as a series of links and paths, where the model is tasked with predicting per path delay and jitter.
• The model has a multi-step message passing routine, first aggregating state information for the links in a given path, and then passing the sequence of link states through a RNN to update the path state.
• The updated path states are then used to update the link states for all paths that utilize a given link.
• This implementation was done as part of a group project for CSE 222 at UC San Diego, where I was responsible for the model creation and training.
  • This repo is a simplified and lightly refactored version of the original project Repo.