SBI_dynamical_mass_estimator

Dynamical mass inference of galaxy clusters using a 3D convolutional neural network

The network architecture & training methodology, mock cluster catalogue and results are detailed in:

"Simulation-based inference of dynamical galaxy cluster masses with 3D convolutional neural networks,"
Doogesh Kodi Ramanah, Radoslaw Wojtak, Nikki Arendse [arXiv:2009.03340]

Note:

1. Network implementation and training routine for the 3D CNN model is in train.py;
2. The notebook dynamical_mass_inference.ipynb provides a stepwise description of the simulation-based inference aspect and illustrates some plots;
3. The mock cluster generated for our project is available here. Please cite the above paper if you make use of our mock cluster catalogue. The script gen_KDE_mock.py shows an example of how to load and text files and generate their 3D KDE representations as employed in our work.