Home
Rodrigo Luger edited this page Sep 4, 2017
·
2 revisions
Main goals for everest3:
- A complete re-write of the code The current code is quite clunky and not that developer-friendly
- Speed up the GP with celerite We will simultaneously optimize the GP and the PLD parameters since celerite is so fast. This will be awesome for stellar rotation studies
- Code should be mission-independent I tried to do this with everest2 but didn't really succeed. It should work for Kepler (and ideally TESS) out-of-the-box
- Code should be less exoplanet-specific Lots of people interested in supernovae/asteroseismology/etc would like to use everest
- Fix overfitting issues We need to investigate what's going on with nPLD, particularly during thruster fires
- Code with Campaigns 16+ in mind The roll will get really bad once the thrusters start sputtering. NKS and I are currently working on this
- Re-think PLD PLD is great but somewhat hacky. Normalizing the pixels by the SAP flux introduces more noise into the regressors and is un-Bayesian. There may be a more elegant (and efficient) way to de-trend