| Name: NIPUN GHANGHAS |
| Affiliation: TATA INSTITUTE OF FUNDAMENTAL RESEARCH |
| Conference ID: ASI2026_787 |
| Title: A Deep Learning Framework for the classification and characterization of solar-like pulsations from PLATO data |
| Abstract Type: Poster |
| Abstract Category: Stars, Interstellar Medium, and Astrochemistry in Milky Way |
| Author(s) and Co-Author(s) with Affiliation: NIPUN GHANGHAS(TATA INSTITUTE OF FUNDAMETAL RESEARCH, Mumbai - 400005, India), Chris Hanson(Center for Space Science, NYUAD Research Institute, New York University Abu Dhabi, UAE), Othman Benomar(Division of Solar and Plasma Astrophysics, NAOJ, Mitaka, Tokyo, Japan) |
| Abstract: The ESA PLATO (PLAnetary Transits and Oscillations of stars) mission is expected to increase the number of well-characterized main-sequence and subgiant stars with detectable solar-like oscillations by more than a factor of 40 compared to the Kepler Legacy sample. This unprecedented data volume necessitates fast, robust, and automated analysis tools capable of extracting key asteroseismic information at scale.
We present a machine-learning-based asteroseismic pipeline designed to process PLATO power spectra and deliver the essential parameters required for both estimating stellar mass and radius and for estimating oscillation mode frequencies, providing reliable initial guesses for detailed peak-bagging analyses. The pipeline is trained on a large synthetic dataset spanning a wide range of stellar and observational conditions, including stellar inclination, signal-to-noise ratio, and global seismic properties such as ν_max and Δν, ensuring robustness across the diverse PLATO target population.
The first stage of the pipeline employs a one-dimensional convolutional neural network (1D-CNN) to automatically classify stars as main-sequence oscillators, subgiants, or non-oscillators. For stars identified as main-sequence solar-like oscillators, the pipeline estimates the global seismic parameters ν_max and Δν, which can be used for determining stellar mass and radius through established scaling relations. In subsequent stages, two-dimensional CNNs are used to extract additional asteroseismic observables, such as the phase offset and small frequency separations, which encode information about stellar interior structure and are needed for estimating individual mode frequencies.
This pipeline provides a scalable and physically motivated framework for exploiting PLATO’s large asteroseismic sample, enabling efficient stellar characterization and supporting both stellar physics and exoplanet science in the PLATO era. |
