Skip to main content
SearchLoginLogin or Signup

A Machine-Learning-Ready Volumetric Active Region Dataset

Presentation #110.06 in the session Data Analysis Techniques Posters.

Published onSep 18, 2023
A Machine-Learning-Ready Volumetric Active Region Dataset

Physics-Informed Neural Networks (PINNs) and other methods for construction or analysis of full coronal magnetic vector fields are becoming more prominent in literature. A need has developed for a large set of simulated active regions for training, validation and testing purposes. We use a state-of-the-art magnetohydrostatic extrapolation method to develop a public dataset of over five thousand data cubes based on the SHARP library, including both plasma-forced and nonlinear force-free models. Each cube resolves the magnetic field vector and plasma forcing at scattered points that are adaptively clustered near the high-flux regions of the domain.We present not only the structure and access of the dataset with emphasis on some example cubes, but also preliminary results on a PINN which was trained on the dataset to extrapolate magnetic fields from photospheric observation.

No comments here