Estimating TESS Stellar Rotation Periods with CNNs
Mission Overview
Estimating TESS stellar rotation periods with CNNs
The NASA TESS mission conducts an all-sky survey searching for exoplanets that transit their host stars. To do so, it collects time series photometry called “light curves” for millions of stars across the sky. These light curves have many science uses besides exoplanets, including stellar rotation. As a star rotates, cool magnetic spots come into and out of view, causing periodic wiggles in the brighness measurements through time. We can therefore use light curves from TESS to infer stellar rotation periods, which are useful for studying stellar magnetism, structure, and ages.
In this tutorial we use a convolutional neural network (CNN) to estimate stellar rotation periods from frequency transforms of TESS light curves. For our training set, we will use the simulations from the MAST High Level Science Product SMARTS. SMARTS combines physically realistic simulations of rotational light curves with real noise and systematics from TESS. This combination allows CNNs to learn the difference between rotation signals and systematics and estimate stellar rotation periods.
Data:The SMARTS HLSP
Notebook: Estimating TESS stellar rotation periods with CNNs
Released: 2025年09月02日
Updated: 2025年09月02日
Tags: neural networks, 2d data, supervised, regression, convolutional neural networks