NERSC / dayabay-learn

Learning to Extract Features from the Daya Bay Reactor Neutrino Experiment

Daya Bay Machine Learning Software

This repository contains a set of neural networks developed to learn more about the data collected by the Daya Bay Reactor Neutrino Experiment.

Instructions:

There are 2 main scripts: a command-line based python script, ibd_ae.py, and an IPython Notebook, ibd_ae.ipynb. There is also a helper script to make t-SNE plots, tsne.py.

To set up the Cori environment, run the following commands

module load python
module load deeplearning 

On other environments, ensure the following dependencies are available:

• python
• theano
• lasagne
• scikit-learn
• numpy/scipy/matplotlib
• h5py

To run the IPython notebook, open it up (e.g. at ipython.nersc.gov) and follow the input prompts. To run the command-line script, you can find help by executing the following command:

python ibd_ae.py --help

There are multiple network architectures located in the networks directory. View the README there for more information.