Skip to content

Chris-Pedersen/LearnableWavelets

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

219 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Learnable Wavelet Networks

Scattering convolutional networks with learnable filter parameters implemented in PyTorch.

Code used in the ICML 2022 Workshop on Machine Learning for Astrophysics workshop paper "Learnable Wavelet Neural Networks for Cosmological Inference". The code has been significantly refactored since this paper but produces consistent results. Nevertheless for completeness, the specific version of the code used to generate results for this paper is here, and requires a custom version of the scattering transform code kymatio at found here due to padding issues in the standard library.

Most of the models code is lifted from https://github.com/bentherien/ParametricScatteringNetworks/, and most of the CAMELs code is lifted from https://camels-multifield-dataset.readthedocs.io/en/latest/.

Set up environment in conda

If using conda, set up an environment in the following way:

conda env create -f environment.yml

Once installed, activate the environment:

conda activate wavelet

and install the learnable_wavelet package via

pip install .

Running example notebook

NB to run the playground/example_notebook.ipynb, first run the following to download the small sample of CAMELs data:

wget -O test_data/maps_Mtot_1k.npy "https://www.dropbox.com/s/oqy3h3l4cdgfdeq/maps_Mtot_1k.npy?dl=0"

About

Learnable wavelet neural networks

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors