Cosmological 21cm signal modelling with tensorflow

Version: 1.0

twentyoneflow allows on to make fast predictions of the 21cm signal using deep neural networks developed and trained using tensorflow. The

An example of using pre-trained models to make predictions can be found in notebooks/example.ipynb.

NOTE: All training of models in this repo was done using the 21cm cosomology training data found at:

https://zenodo.org/record/3532141#.XcSg7NHLc5k

Furthermore, example.ipynb uses the 21cm cosmology test data available at:

https://zenodo.org/record/3530920#.XcRlKjP7QuU

These two datasets were obtained from randomly subsampling a dataset which forms their union.

The pre-trained networks can be found in saved_models/keras/, while the preprocessing scalers are in saved_models/scalers/.

An example script for training your own model using tensorflow/keras can be found in scripts/keras_model_template.py, while sample architectures can be found in twentyoneflow/keras_models.py. Similarly a template for sklearn networks can be found in scripts/sklearn_model_template.py.

This package was developed and tested using python 2.7, but should be compatible with newer versions such as python 3.6.

Coming soon.

No installation of twentyoneflow is strictly necessary, one can simply clone the repo:

git clone https://github.com/SuperKam91/twentyoneflow.git,

and run their code from there, or add the location of the directory to their PYTHONPATH environment variable. Note however that one must ensure that all the required packages are installed by running pip install -r requirements.txt from the root the directory.

Alternatively, one can run

python setup.py install

or

pip install .

from the root of the repo to install the base requirements (i.e. those in requirements.txt) as well as the twentyoneflow package.

If the previous method doesn't work, first run python setup.py bdist_wheel followed by python -m pip install dist/<wheel file name>.whl where <wheel file name> will depend on the version of python used when running setup.py and the version number contained within the file.

Want to contribute to twentyoneflow? There are two main you can contribute via the GitHub repository:

Open an issue to report bugs or to propose new features.

Pull requests are very welcome. Note that if you are going to propose drastic changes, please open an issue for discussion first, to make sure that your proposed change will be accepted before you spend effort implementing it.