stela-toolkit 0.3.0

A Python package for interpolating astrophysical light curves using Gaussian Processes (more ML models to come!) in order to compute frequency-domain and standard time domain data products.

Logo design by Elizabeth Jarboe

STELA (Sampling Time for Even Lightcurve Analysis) Toolkit is a Python package for interpolating astrophysical light curves using Gaussian Processes (more ML models to come!) in order to compute frequency-domain and standard time domain data products.

---

Documentation

Full documentation is available at:
https://collinlewin.github.io/stela-toolkit/

The documentation includes:

• Overview of all features
• Tutorial notebook with GP modeling and lag analysis
• Gaussian Process guide with detailed Bayesian background
• Module reference for all functions and classes

---

Features

• Gaussian Process regression for all ML skill levels! Built-in normality testing and transformations to help reach normality, kernel selection, hyperparameter training, and much more.
• Frequency-domain products: power and cross spectra, coherence, lag-frequency and lag-energy spectra
• Time-domain lag analysis using ICCF and GP-based cross-correlation
• Simulation of synthetic light curves with custom underlying power spectra and injected lags
• Preprocessing: outlier removal methods, polynomial detrending, trimming, etc.
• Convenient plotting in every class using .plot()

---

Requirements

STELA Toolkit requires Python ≥ 3.8 and the following core packages:

• numpy ≥ 2.0
• scipy ≥ 1.7
• matplotlib ≥ 3.5
• astropy ≥ 5.0
• torch ≥ 1.10
• gpytorch ≥ 1.9
• statsmodels ≥ 0.13

You can install all dependencies automatically when installing the package (see below).

---

Installation

You can install the package using pip:

pip install stela-toolkit

If you are installing directly from the GitHub repository:

git clone https://github.com/collinlewin/STELA-Toolkit.git
cd STELA-Toolkit
pip install .

You can verify the installation by running:

import stela_toolkit
print(stela_toolkit.__version__)

If this runs without an error, you're good to go!