eurydice 0.3

A package for evaluating radival velocity signal models using cross-validation methods and Gaussian processes.

EURYDICE: EvalUating Radial velocitY moDels usIng Cross-validation and Gaussian procEsses

A WIP to create a package that will perform cross-validations for radial velocity signal models with Gaussian processes 💃🕺

Last Updated: July 24th 2024

Current Version: 0.3

v.0.2 Updates (July 10th 2024):

• autoformatting
• splitting code into modules: kepler (for keplerian helper functions), plot (to handle plotting), and CV (holding the CrossValidation object + all GPR code).
• updating some old error messages in kepler.calc_keplerian_signal function
• added a default kernel for the CrossValidation object to use if a predefined one is not passed
• updating plotting functions to return a matplotlib.plyplot.Figure
• fitting Gaussians to the residuals on the histogram plot used for CV
• added functionality for the CrossValidation object to calculate and utilize an N body keplerian mean function

To-Do:

• adding compatibility with pre-existing GP packages (george, tinygp, celerite): the gist is probably to let code read if the kernel function its passed is a certain object (e.g a george kernel object) and let the code run the GPR and CV using that package's methods
• writing a tutorial on how to use code with juypter notebook
• combing through existing code to improve lacking documentation and find more appropriate places to include error messages
• modifying split function in CrossValidation to let users have a choice as to split data randomly or not (e.g whether to split data 80/20 randomly for training or to split the data such that the first 80 data points chronologically are used to train the model and the last 20 used to assess predictability)
• writing more tests: how to write tests for GP_predict, run_CV and plotting functions? seems too convoluted for my pea brain at the moment 🫨