| _version_ | 1866901058966519808 |
|---|---|
| author | Ghanghas, Nipun |
| author_facet | Ghanghas, Nipun |
| contents | <p><span>Asteroseismology enables precise inference of stellar properties by analyzing oscillation modes in stellar light curves. Missions such as Kepler, K2, and TESS have provided observations of hundreds of thousands of pulsating stars across a wide range of evolutionary stages. These datasets will expand further with upcoming missions like PLATO and Roman. Their scale calls for fast, robust, and automated methods to extract asteroseismic parameters.<br><br>We present machine learning (ML) frameworks developed to analyze red giants observed by TESS and K2, as well as main-sequence and subgiant stars expected from PLATO. For red giants, we use dedicated models to infer the large frequency separation (Δν) and the frequency of maximum power (νmax) from one-month segments of TESS data. For K2 red giants, we extend the analysis to also infer the gravity-mode period spacing (ΔΠ1). These models are validated against benchmark values from the literature using one- and three-month segments of Kepler data.<br><br>Notably, for K2, we have inferred ΔΠ1 for approximately 200 young red giants—for the first time—showing good agreement with the well-known Δν–ΔΠ1 degenerate sequence.<br><br>In parallel, we develop a pipeline for main-sequence stars using synthetic data representative of PLATO observations. A convolutional neural network first classifies stars as main-sequence, subgiant, or non-oscillators. For oscillating stars, a specialized 2D CNN infers νmax, Δν, the oscillation phase offset (epsilon), linewidth at νmax, and the average small frequency spacing.<br><br>Together, these pipelines enable automated asteroseismic analysis across stellar types and missions, laying the groundwork for future population-level studies in the era of precision space-based photometry.</span></p> |
| format | Recurso digital |
| id | zenodo_https___doi_org_10_5281_zenodo_18129777 |
| institution | Zenodo |
| language | |
| publishDate | 2026 |
| publisher | Zenodo |
| record_format | zenodo |
| spellingShingle | From Kepler to PLATO: ML-Based Asteroseismic Inference Across Stellar Populations Ghanghas, Nipun <p><span>Asteroseismology enables precise inference of stellar properties by analyzing oscillation modes in stellar light curves. Missions such as Kepler, K2, and TESS have provided observations of hundreds of thousands of pulsating stars across a wide range of evolutionary stages. These datasets will expand further with upcoming missions like PLATO and Roman. Their scale calls for fast, robust, and automated methods to extract asteroseismic parameters.<br><br>We present machine learning (ML) frameworks developed to analyze red giants observed by TESS and K2, as well as main-sequence and subgiant stars expected from PLATO. For red giants, we use dedicated models to infer the large frequency separation (Δν) and the frequency of maximum power (νmax) from one-month segments of TESS data. For K2 red giants, we extend the analysis to also infer the gravity-mode period spacing (ΔΠ1). These models are validated against benchmark values from the literature using one- and three-month segments of Kepler data.<br><br>Notably, for K2, we have inferred ΔΠ1 for approximately 200 young red giants—for the first time—showing good agreement with the well-known Δν–ΔΠ1 degenerate sequence.<br><br>In parallel, we develop a pipeline for main-sequence stars using synthetic data representative of PLATO observations. A convolutional neural network first classifies stars as main-sequence, subgiant, or non-oscillators. For oscillating stars, a specialized 2D CNN infers νmax, Δν, the oscillation phase offset (epsilon), linewidth at νmax, and the average small frequency spacing.<br><br>Together, these pipelines enable automated asteroseismic analysis across stellar types and missions, laying the groundwork for future population-level studies in the era of precision space-based photometry.</span></p> |
| title | From Kepler to PLATO: ML-Based Asteroseismic Inference Across Stellar Populations |
| url | https://doi.org/10.5281/zenodo.18129777 |