ARG

Artificially Redshifting Galaxies with a neural network

Instructions for generating galaxy images and training the neural network:

Dependancies:

Numpy
Scipy
Astropy
H5PY
Six
Sklearn
Tensorflow

Clone and install SMPY --> https://github.com/dunkenj/smpy

git clone git://github.com/dunkenj/smpy.git

python setup.py install

Run the scripts

Run candels_example.py from the scripts folder of SMPY to generate the galaxy SED information used by galaxies.py.

Run galaxies.py from the scripts folder to generate n simulated galaxy images.

Run redshifting.py from the scripts folder to apply observational effects to the n galaxy images.

Run Galaxies_VAE.pyfrom the scripts folder to train the network and produce image reconstructions, a 2D plot in latent space, and plot of model loss on training and validation sets.

Additional scripts

galaxies2.py can be used to generate compound galaxy images.

galaxies_asym.py can be used to generate asymmetric galaxy images.

Note

The outputs of galaxies.py and redshifting.py will be .npy files.

testplot.py shows how the numpy arrays can be read and displayed.