Spherinator and HiPSter are tools to provide much needed explorative access and visualization for multimodal data from extremely large astrophysical datasets, ranging from exascale cosmological simulations to multi-billion object observational galaxy surveys. Spherinator uses dimensionality reduction to learn a low-dimensional representation of galaxy structure, and HiPSter creates a interactive hierarchical spherical vizualization of the entire dataset. They currently support multichannel maps or images as input. Spherinator uses PyTorch Lightning to implement a convolutional neural network (CNN) based variational autoencoder (VAE) with a spherical latent space.
HiPSter creates HiPS tilings and a catalog which can be visualized interactively on the surface of a sphere using Aladin Lite.
pip install spherinatorFunded by the European Union. This work has received funding from the European High-Performance Computing Joint Undertaking (JU) and Belgium, Czech Republic, France, Germany, Greece, Italy, Norway, and Spain under grant agreement No 101093441.
Views and opinions expressed are however those of the author(s) only and do not necessarily reflect those of the European Union or the European High Performance Computing Joint Undertaking (JU) and Belgium, Czech Republic, France, Germany, Greece, Italy, Norway, and Spain. Neither the European Union nor the granting authority can be held responsible for them.
This project is licensed under the Apache-2.0 License.
If you use Spherinator & HiPSter in your research, we provide a citation to use:
@article{Polsterer_Spherinator_and_HiPSter_2024,
author = {Polsterer, Kai Lars and Doser, Bernd and Fehlner, Andreas and Trujillo-Gomez, Sebastian},
title = {{Spherinator and HiPSter: Representation Learning for Unbiased Knowledge Discovery from Simulations}},
url = {https://arxiv.org/abs/2406.03810},
doi = {10.48550/arXiv.2406.03810},
year = {2024}
}