Standard methods for inferring planetary characteristics from exoplanets’ atmospheric spectra are slow. We present a fast machine learning method: a deep ensemble of convolutional neural networks (CNNs) that outputs mixtures of normal distributions for planetary characteristics. The architecture of our CNN was inspired by VGG networks. We train each CNN with Kullback–Leibler divergence as its loss function on simulated exoplanet’ atmospheric spectra and their auxiliary data from Ariel Data Challenge.
The Python script named ariel.py trains models, and the Jupyter notebook named test.ipynb generates the required outputs.
See slides.pdf for some information.
Python 3.9.6, CUDA 11.3.1, and PyTorch 1.12.1.
$ python -m venv venv
$ source venv/bin/activate
$ pip install torch==1.12.1+cu113 --extra-index-url https://download.pytorch.org/whl/cu113
$ pip install -r requirements.txt
Data are available on Zenodo, at https://doi.org/10.5281/zenodo.6770103.
Concretely, it is the ZIP file Level2Data.zip.
Ondřej Podsztavek (podszond@fit.cvut.cz)
Lakshminarayanan, B., Pritzel, A., Blundell, C., 2017. Simple and Scalable Predictive Uncertainty Estimation using Deep Ensembles. In: Guyon, I., et al. (Eds.), Advances in Neural Information Processing Systems 30.
Simonyan, K., Zisserman, A., 2015. Very deep convolutional networks for large-scale image recognition. arXiv:1409.1556.