Repository for software from the paper: "Extracting the cold neutral medium from HI emission with deep learning: Implications for Galactic foregrounds at high latitude" by Claire E. Murray, J.E.G. Peek and Chang-Goo Kim (2020 ApJ submitted).
With this notebook you should be able to:
- access remote data, including training and test data, as well as observed catalog information and CNN maps
- build and train a 1D CNN to predict f_CNM and R_HI from 21cm brightness temperature spectra (TB(v))
- assess what the network is learning by computing the saliency for the output predictions
- compare the input values with the network predictions
- re-create Figures 6, 8, 9, and 14
The following dependencies are used to run the Jupyter Notebook (including latest tested versions):
- keras 2.3.1
- tensorflow 2.1.0
- keras-vis
- numpy 1.18.2
- astropy 4.0.1
- tqdm 4.32.2
- pickle
- scipy 1.4.1
- sklearn 0.19.1
A conda environment can be created using the env.yml
file:
conda env create -f env.yml
conda activate cnn_cnm
pip install -I git+https://github.com/raghakot/keras-vis.git
Data in other formats/projections available upon request (please contact @cmurray-astro).