The L-retrieval project develops a new way to retrieve exoplanet characteristic from telescope data. We demonstrated that one can extract vital atmospheric information directly from the instrument light-curves, without extracting their wave-length dependent transit depths.
The Movies Project
This project has an individual showcase page, not just a direct link to the project site or repo. Now you have more space to describe your awesome project!
Black Box Sensitivity Analysis
Sensitivity analysis for machine learning and deep learning models. This perturbation based analysis helps to quantify the relative importance of each input features w.r.t. the target(s). We developed this technique in order to verify whether our trained model aligns with our physical intuition, in our case exoplanet spectroscopy. The full details of our project can be found here.
Direct Imaging is one of the widely used techniques in discovering new exoplanet and characterising existing ones. This project provides a Proof-of-Concept study to access the feasbility of applying state-of-the-art deep learning techniques to help discover exoplanets from existing datasets.
This work analysed transit signal coming from WASP-96 b, a hot Neptune class exoplanet. After detrending raw-lightcurves from HST grism data, we discovered water signature on the planet's atmosphere. However, we also realised an abrupt discrepancy between ground based and space based instruments. Click to learn more.