Probabilistic Graphical Transportation Networks

This is the final project of the course 10-708 Probablistic Graphical Models at Carnegie Mellon University. It studies the use of Restricted Boltzmann Machines, Deep Belief Network and Variational AutoEncoders to estimate Origin-Destination matrices. The analyses were performed using synthetic data and the models were implemented with TensorFlow primarily.

Development setup

1. Clone the repository
2. Create virtual environment (e.g. "venv") python3 -m venv venv
3. Activate virtual environment source venv/bin/activate
4. Install the development dependencies: pip install -r requirements.txt
5. Run python3 main.py

Visualizations