Multi-output Deep Learning Framework for Solar Atmospheric Parameters Inferring from Stokes Profiles
01 Jan 2022-pp 299-307
About: The article was published on 2022-01-01 and is currently open access. It has received 2 citations till now.
Citations
More filters
26 Oct 2022
TL;DR: This paper provides end-to-end inversion codes based on the simple Milne-Eddington model of the stellar atmosphere and deep neural networks to both parameter estimation and their uncertainty intervals and demonstrates that the proposed architecture provides high accuracy of results, including a reliable uncertainty estimation, even in the multidimensional case.
Abstract: Magnetic fields are responsible for a multitude of Solar phenomena, including such destructive events as solar flares and coronal mass ejections, with the number of such events rising as we approach the peak of the 11-year solar cycle, in approximately 2025. High-precision spectropolarimetric observations are necessary to understand the variability of the Sun. The field of quantitative inference of magnetic field vectors and related solar atmospheric parameters from such observations has long been investigated. In recent years, very sophisticated codes for spectropolarimetric observations have been developed. Over the past two decades, neural networks have been shown to be a fast and accurate alternative to classic inversion technique methods. However, most of these codes can be used to obtain point estimates of the parameters, so ambiguities, the degenera-cies, and the uncertainties of each parameter remain uncovered. In this paper, we provide end-to-end inversion codes based on the simple Milne-Eddington model of the stellar atmosphere and deep neural networks to both parameter estimation and their uncertainty intervals. The proposed framework is designed in such a way that it can be expanded and adapted to other atmospheric models or combinations of them. Additional information can also be incorporated di-rectly into the model. It is demonstrated that the proposed architecture provides high accuracy of results, including a reliable uncertainty estimation, even in the multidimensional case. The models are tested using simulation and real data samples.
References
More filters
Posted Content•
TL;DR: This article seeks to help ML practitioners apply MTL by shedding light on how MTL works and providing guidelines for choosing appropriate auxiliary tasks, particularly in deep neural networks.
Abstract: Multi-task learning (MTL) has led to successes in many applications of machine learning, from natural language processing and speech recognition to computer vision and drug discovery. This article aims to give a general overview of MTL, particularly in deep neural networks. It introduces the two most common methods for MTL in Deep Learning, gives an overview of the literature, and discusses recent advances. In particular, it seeks to help ML practitioners apply MTL by shedding light on how MTL works and providing guidelines for choosing appropriate auxiliary tasks.
2,202 citations
TL;DR: In this paper, the authors describe the implementation and main properties of a new inversion code for the polarized radiative transfer equation (VFISV: Very Fast Inversion of the Stokes Vector).
Abstract: In this paper we describe in detail the implementation and main properties of a new inversion code for the polarized radiative transfer equation (VFISV: Very Fast Inversion of the Stokes Vector). VFISV will routinely analyze pipeline data from the Helioseismic and Magnetic Imager (HMI) on-board of the Solar Dynamics Observatory (SDO). It will provide full-disk maps (4096×4096 pixels) of the magnetic field vector on the Solar Photosphere every ten minutes. For this reason VFISV is optimized to achieve an inversion speed that will allow it to invert sixteen million pixels every ten minutes with a modest number (approx. 50) of CPUs. Here we focus on describing a number of important details, simplifications and tweaks that have allowed us to significantly speed up the inversion process. We also give details on tests performed with data from the spectropolarimeter on-board of the Hinode spacecraft.
358 citations
348 citations
TL;DR: In this paper, the authors revisited the foundations of inversion techniques with an emphasis on making explicit the many assumptions underlying each of them, including the physics assumed to prevail both on the formation of the spectral line Stokes profiles and on their detection with the instrument.
Abstract: Since the early 1970s, inversion techniques have become the most useful tool for inferring the magnetic, dynamic, and thermodynamic properties of the solar atmosphere Inversions have been proposed in the literature with a sequential increase in model complexity: astrophysical inferences depend not only on measurements but also on the physics assumed to prevail both on the formation of the spectral line Stokes profiles and on their detection with the instrument Such an intrinsic model dependence makes it necessary to formulate specific means that include the physics in a properly quantitative way The core of this physics lies in the radiative transfer equation (RTE), where the properties of the atmosphere are assumed to be known while the unknowns are the four Stokes profiles The solution of the (differential) RTE is known as the direct or forward problem From an observational point of view, the problem is rather the opposite: the data are made up of the observed Stokes profiles and the unknowns are the solar physical quantities Inverting the RTE is therefore mandatory Indeed, the formal solution of this equation can be considered an integral equation The solution of such an integral equation is called the inverse problem Inversion techniques are automated codes aimed at solving the inverse problem The foundations of inversion techniques are critically revisited with an emphasis on making explicit the many assumptions underlying each of them
94 citations
TL;DR: It is found that multi-task learning typically does not improve performance for a user-defined combination of tasks, and requires careful selection of both task pairs and weighting strategies to equal or exceed the performance of single task learning.
Abstract: With the success of deep learning in a wide variety of areas, many deep multi-task learning (MTL) models have been proposed claiming improvements in performance obtained by sharing the learned structure across several related tasks. However, the dynamics of multi-task learning in deep neural networks is still not well understood at either the theoretical or experimental level. In particular, the usefulness of different task pairs is not known a priori. Practically, this means that properly combining the losses of different tasks becomes a critical issue in multi-task learning, as different methods may yield different results. In this paper, we benchmarked different multi-task learning approaches using shared trunk with task specific branches architecture across three different MTL datasets. For the first dataset, i.e. Multi-MNIST (Modified National Institute of Standards and Technology database), we thoroughly tested several weighting strategies, including simply adding task-specific cost functions together, dynamic weight average (DWA) and uncertainty weighting methods each with various amounts of training data per-task. We find that multi-task learning typically does not improve performance for a user-defined combination of tasks. Further experiments evaluated on diverse tasks and network architectures on various datasets suggested that multi-task learning requires careful selection of both task pairs and weighting strategies to equal or exceed the performance of single task learning.
73 citations