Supervised neural networks for helioseismic ring-diagram inversions

Research output: Contribution to journalArticle

Abstract

Context. The inversion of ring fit parameters to obtain subsurface flow maps in ring-diagram analysis for eight years of SDO observations is computationally expensive, requiring ∼3200 CPU hours. Aims. In this paper we apply machine-learning techniques to the inversion step of the ring diagram pipeline in order to speed up the calculations. Specifically, we train a predictor for subsurface flows using the mode fit parameters and the previous inversion results to replace future inversion requirements. Methods. We utilize artificial neural networks (ANNs) as a supervised learning method for predicting the flows in 15° ring tiles. We discuss each step of the proposed method to determine the optimal approach. In order to demonstrate that the machine-learning results still contain the subtle signatures key to local helioseismic studies, we use the machine-learning results to study the recently discovered solar equatorial Rossby waves. Results. The ANN is computationally efficient, able to make future flow predictions of an entire Carrington rotation in a matter of seconds, which is much faster than the current ∼31 CPU hours. Initial training of the networks requires ∼3 CPU hours. The trained ANN can achieve a rms error equal to approximately half that reported for the velocity inversions, demonstrating the accuracy of the machine learning (and perhaps the overestimation of the original errors from the ring-diagram pipeline). We find the signature of equatorial Rossby waves in the machine-learning flows covering six years of data, demonstrating that small-amplitude signals are maintained. The recovery of Rossby waves in the machine-learning flow maps can be achieved with only one Carrington rotation (27.275 days) of training data. Conclusions. We show that machine learning can be applied to and perform more efficiently than the current ring-diagram inversion. The computation burden of the machine learning includes 3 CPU hours for initial training, then around 10 -4 CPU hours for future predictions.

Original languageEnglish (US)
Article numberA124
JournalAstronomy and Astrophysics
Volume622
DOIs
StatePublished - Feb 1 2019

Fingerprint

machine learning
diagram
diagrams
inversions
rings
Rossby wave
planetary waves
artificial neural network
solar rotation
equatorial wave
education
subsurface flow
predictions
signatures
ring currents
inversion
tiles
prediction
learning
train

Keywords

  • Methods: numerical
  • Sun: helioseismology
  • Sun: interior
  • Sun: oscillations

ASJC Scopus subject areas

  • Astronomy and Astrophysics
  • Space and Planetary Science

Cite this

Supervised neural networks for helioseismic ring-diagram inversions. / Alshehhi, Rasha; Hanson, Chris S.; Gizon, Laurent; Hanasoge, Shravan.

In: Astronomy and Astrophysics, Vol. 622, A124, 01.02.2019.

Research output: Contribution to journalArticle

@article{82cadf2d41834d348d57d1da1fd65de3,
title = "Supervised neural networks for helioseismic ring-diagram inversions",
abstract = "Context. The inversion of ring fit parameters to obtain subsurface flow maps in ring-diagram analysis for eight years of SDO observations is computationally expensive, requiring ∼3200 CPU hours. Aims. In this paper we apply machine-learning techniques to the inversion step of the ring diagram pipeline in order to speed up the calculations. Specifically, we train a predictor for subsurface flows using the mode fit parameters and the previous inversion results to replace future inversion requirements. Methods. We utilize artificial neural networks (ANNs) as a supervised learning method for predicting the flows in 15° ring tiles. We discuss each step of the proposed method to determine the optimal approach. In order to demonstrate that the machine-learning results still contain the subtle signatures key to local helioseismic studies, we use the machine-learning results to study the recently discovered solar equatorial Rossby waves. Results. The ANN is computationally efficient, able to make future flow predictions of an entire Carrington rotation in a matter of seconds, which is much faster than the current ∼31 CPU hours. Initial training of the networks requires ∼3 CPU hours. The trained ANN can achieve a rms error equal to approximately half that reported for the velocity inversions, demonstrating the accuracy of the machine learning (and perhaps the overestimation of the original errors from the ring-diagram pipeline). We find the signature of equatorial Rossby waves in the machine-learning flows covering six years of data, demonstrating that small-amplitude signals are maintained. The recovery of Rossby waves in the machine-learning flow maps can be achieved with only one Carrington rotation (27.275 days) of training data. Conclusions. We show that machine learning can be applied to and perform more efficiently than the current ring-diagram inversion. The computation burden of the machine learning includes 3 CPU hours for initial training, then around 10 -4 CPU hours for future predictions.",
keywords = "Methods: numerical, Sun: helioseismology, Sun: interior, Sun: oscillations",
author = "Rasha Alshehhi and Hanson, {Chris S.} and Laurent Gizon and Shravan Hanasoge",
year = "2019",
month = "2",
day = "1",
doi = "10.1051/0004-6361/201834237",
language = "English (US)",
volume = "622",
journal = "Astronomy and Astrophysics",
issn = "0004-6361",
publisher = "EDP Sciences",

}

TY - JOUR

T1 - Supervised neural networks for helioseismic ring-diagram inversions

AU - Alshehhi, Rasha

AU - Hanson, Chris S.

AU - Gizon, Laurent

AU - Hanasoge, Shravan

PY - 2019/2/1

Y1 - 2019/2/1

N2 - Context. The inversion of ring fit parameters to obtain subsurface flow maps in ring-diagram analysis for eight years of SDO observations is computationally expensive, requiring ∼3200 CPU hours. Aims. In this paper we apply machine-learning techniques to the inversion step of the ring diagram pipeline in order to speed up the calculations. Specifically, we train a predictor for subsurface flows using the mode fit parameters and the previous inversion results to replace future inversion requirements. Methods. We utilize artificial neural networks (ANNs) as a supervised learning method for predicting the flows in 15° ring tiles. We discuss each step of the proposed method to determine the optimal approach. In order to demonstrate that the machine-learning results still contain the subtle signatures key to local helioseismic studies, we use the machine-learning results to study the recently discovered solar equatorial Rossby waves. Results. The ANN is computationally efficient, able to make future flow predictions of an entire Carrington rotation in a matter of seconds, which is much faster than the current ∼31 CPU hours. Initial training of the networks requires ∼3 CPU hours. The trained ANN can achieve a rms error equal to approximately half that reported for the velocity inversions, demonstrating the accuracy of the machine learning (and perhaps the overestimation of the original errors from the ring-diagram pipeline). We find the signature of equatorial Rossby waves in the machine-learning flows covering six years of data, demonstrating that small-amplitude signals are maintained. The recovery of Rossby waves in the machine-learning flow maps can be achieved with only one Carrington rotation (27.275 days) of training data. Conclusions. We show that machine learning can be applied to and perform more efficiently than the current ring-diagram inversion. The computation burden of the machine learning includes 3 CPU hours for initial training, then around 10 -4 CPU hours for future predictions.

AB - Context. The inversion of ring fit parameters to obtain subsurface flow maps in ring-diagram analysis for eight years of SDO observations is computationally expensive, requiring ∼3200 CPU hours. Aims. In this paper we apply machine-learning techniques to the inversion step of the ring diagram pipeline in order to speed up the calculations. Specifically, we train a predictor for subsurface flows using the mode fit parameters and the previous inversion results to replace future inversion requirements. Methods. We utilize artificial neural networks (ANNs) as a supervised learning method for predicting the flows in 15° ring tiles. We discuss each step of the proposed method to determine the optimal approach. In order to demonstrate that the machine-learning results still contain the subtle signatures key to local helioseismic studies, we use the machine-learning results to study the recently discovered solar equatorial Rossby waves. Results. The ANN is computationally efficient, able to make future flow predictions of an entire Carrington rotation in a matter of seconds, which is much faster than the current ∼31 CPU hours. Initial training of the networks requires ∼3 CPU hours. The trained ANN can achieve a rms error equal to approximately half that reported for the velocity inversions, demonstrating the accuracy of the machine learning (and perhaps the overestimation of the original errors from the ring-diagram pipeline). We find the signature of equatorial Rossby waves in the machine-learning flows covering six years of data, demonstrating that small-amplitude signals are maintained. The recovery of Rossby waves in the machine-learning flow maps can be achieved with only one Carrington rotation (27.275 days) of training data. Conclusions. We show that machine learning can be applied to and perform more efficiently than the current ring-diagram inversion. The computation burden of the machine learning includes 3 CPU hours for initial training, then around 10 -4 CPU hours for future predictions.

KW - Methods: numerical

KW - Sun: helioseismology

KW - Sun: interior

KW - Sun: oscillations

UR - http://www.scopus.com/inward/record.url?scp=85061383212&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85061383212&partnerID=8YFLogxK

U2 - 10.1051/0004-6361/201834237

DO - 10.1051/0004-6361/201834237

M3 - Article

AN - SCOPUS:85061383212

VL - 622

JO - Astronomy and Astrophysics

JF - Astronomy and Astrophysics

SN - 0004-6361

M1 - A124

ER -