Searcharxiv⌕ Search

arXiv subjects

Xudong Sun

Publications and source records attributed to Xudong Sun.

At least 73 records · Page 4Linked to original sources
arXiv2019

High Dimensional Restrictive Federated Model Selection with multi-objective Bayesian Optimization over shifted distributions

A novel machine learning optimization process coined Restrictive Federated Model Selection (RFMS) is proposed under the scenario, for example, when data from healthcare units can not leave the site it is situated on and it is forbidden to carry out training algorithms on remote data sites due to either technical or privacy and trust concerns. To carry out a clinical research under this scenario, an analyst could train a machine learning model only on local data site, but it is still possible to execute a statistical query at a certain cost in the form of sending a machine learning model to some of the remote data sites and get the performance measures as feedback, maybe due to prediction being usually much cheaper. Compared to federated learning, which is optimizing the model parameters directly by carrying out training across all data sites, RFMS trains model parameters only on one local data site but optimizes hyper-parameters across other data sites jointly since hyper-parameters play an important role in machine learning performance. The aim is to get a Pareto optimal model with respective to both local and remote unseen prediction losses, which could generalize well across data sites. In this work, we specifically consider high dimensional data with shifted distributions over data sites. As an initial investigation, Bayesian Optimization especially multi-objective Bayesian Optimization is used to guide an adaptive hyper-parameter optimization process to select models under the RFMS scenario. Empirical results show that solely using the local data site to tune hyper-parameters generalizes poorly across data sites, compared to methods that utilize the local and remote performances. Furthermore, in terms of dominated hypervolumes, multi-objective Bayesian Optimization algorithms show increased performance across multiple data sites among other candidates.

cs.LG↗
arXiv2019

ReinBo: Machine Learning pipeline search and configuration with Bayesian Optimization embedded Reinforcement Learning

Machine learning pipeline potentially consists of several stages of operations like data preprocessing, feature engineering and machine learning model training. Each operation has a set of hyper-parameters, which can become irrelevant for the pipeline when the operation is not selected. This gives rise to a hierarchical conditional hyper-parameter space. To optimize this mixed continuous and discrete conditional hierarchical hyper-parameter space, we propose an efficient pipeline search and configuration algorithm which combines the power of Reinforcement Learning and Bayesian Optimization. Empirical results show that our method performs favorably compared to state of the art methods like Auto-sklearn , TPOT, Tree Parzen Window, and Random Search.

cs.LG↗
arXiv2018

Effect of transport coefficients on excitation of flare-induced standing slow-mode waves in coronal loops

Standing slow-mode waves have been recently observed in flaring loops by the Atmospheric Imaging Assembly (AIA) of the Solar Dynamics Observatory (SDO). By means of the coronal seismology technique transport coefficients in hot ($\sim$10 MK) plasma were determined by Wang et al.(2015, Paper I), revealing that thermal conductivity is nearly suppressed and compressive viscosity is enhanced by more than an order of magnitude. In this study we use 1D nonlinear MHD simulations to validate the predicted results from the linear theory and investigate the standing slow-mode wave excitation mechanism. We first explore the wave trigger based on the magnetic field extrapolation and flare emission features. Using a flow pulse driven at one footpoint we simulate the wave excitation in two types of loop models: model 1 with the classical transport coefficients and model 2 with the seismology-determined transport coefficients. We find that model 2 can form the standing wave pattern (within about one period) from initial propagating disturbances much faster than model 1, in better agreement with the observations. Simulations of the harmonic waves and the Fourier decomposition analysis show that the scaling law between damping time ($τ$) and wave period ($P$) follows $τ\propto{P^2}$ in model 2, while $τ\propto{P}$ in model 1. This indicates that the largely enhanced viscosity efficiently increases the dissipation of higher harmonic components, favoring the quick formation of the fundamental standing mode. Our study suggests that observational constraints on the transport coefficients are important in understanding both, the wave excitation and damping mechanisms.

astro-ph.SR↗
arXiv2018

Towards a Quantitative Comparison of Magnetic Field Extrapolations and Observed Coronal Loops

It is widely believed that loops observed in the solar atmosphere trace out magnetic field lines. However, the degree to which magnetic field extrapolations yield field lines that actually do follow loops has yet to be studied systematically. In this paper we apply three different extrapolation techniques - a simple potential model, a NLFF model based on photospheric vector data, and a NLFF model based on forward fitting magnetic sources with vertical currents - to 15 active regions that span a wide range of magnetic conditions. We use a distance metric to assess how well each of these models is able to match field lines to the 12,202 loops traced in coronal images. These distances are typically 1-2". We also compute the misalignment angle between each traced loop and the local magnetic field vector, and find values of 5-12$^\circ$. We find that the NLFF models generally outperform the potential extrapolation on these metrics, although the differences between the different extrapolations are relatively small. The methodology that we employ for this study suggests a number of ways that both the extrapolations and loop identification can be improved.

astro-ph.SR↗
arXiv2018

High resolution observations of sympathetic eruptions by NVST

We investigate two sympathetic filament eruptions observed by the New Vacuum Solar Telescope (NVST) on 2015 October 15. The full picture of the eruptions is obtained from the corresponding SDO/AIA observations. The two filaments start from active region NOAA 12434 in the north and end in one large quiescent filament channel in the south. The left filament erupts firstly, followed by the right filament eruption about 10 minutes later. Clear twist structure and rotating motion are observed in both filaments during the eruption. Both eruptions are failed, since the filaments firstly rise up, then flow towards the south and merge into the southern large quiescent filament. We also observe repeating activations of mini filaments below the right filament after its eruption. Using magnetic field models constructed based on SDO/HMI magnetograms by flux rope insertion method, we find that the left filament eruption is likely to be triggered by kink instability, while weakening of overlying magnetic fields due to magnetic reconnection at an X-point between the two filament systems might play an important role in the onset of the right filament eruption.

astro-ph.SR↗
arXiv2018

Polar Field Correction for HMI Line-of-Sight Synoptic Data

This document provides some technical notes on the polar field correction scheme for the HMI synoptic maps and daily updated synchronic frames. It is intended as a reference for the new data products and for some minor updates on our previous scheme for MDI (Sun et al. 2011).

astro-ph.SR↗
arXiv2017

Electric-Current Neutralization, Magnetic Shear, and Eruptive Activity in Solar Active Regions

The physical conditions that determine whether or not solar active regions (ARs) produce strong flares and coronal mass ejections (CMEs) are not yet well understood. Here we investigate the association between electric-current neutralization, magnetic shear along polarity inversion lines (PILs), and eruptive activity in four ARs; two emerging and two well-developed ones. We find that the CME-producing ARs are characterized by a strongly non-neutralized total current, while the total current in the ARs that did not produce CMEs is almost perfectly neutralized. The difference in the PIL-shear between these two groups is much less pronounced, which suggests that the degree of current-neutralization may serve as a better proxy for assessing the ability of ARs to produce CMEs.

astro-ph.SR↗
arXiv2017

A Database of Flare Ribbon Properties From Solar Dynamics Observatory I: Reconnection Flux

We present a database of 3137 solar flare ribbon events corresponding to every flare of GOES class C1.0 and greater within 45 degrees from the central meridian, from April 2010 until April 2016, observed by the \emph{Solar Dynamics Observatory}. For every event in the database, we compare the GOES peak X-ray flux with corresponding active-region and flare-ribbon properties. We find that while the peak X-ray flux is not correlated with the active region unsigned magnetic flux, it is strongly correlated with the flare ribbon reconnection flux, flare ribbon area, and the fraction of active region flux that undergoes reconnection. We find the relationship between the peak X-ray flux and the flare ribbon reconnection flux to be $I_\mathrm{X,peak} \propto Φ_\mathrm{ribbon}^{1.5}$. This scaling law is consistent with earlier hydrodynamic simulations of impulsively heated flare loops. Using the flare reconnection flux as a proxy for the total released flare energy $E$, we find that the occurrence frequency of flare energies follows a power-law dependence: $dN/dE \propto E^{-1.6}$ for $10^{31}<E<10^{33}$ erg, consistent with earlier studies of solar and stellar flares. The database is available online and can be used for future quantitative studies of flares.

astro-ph.SR↗
arXiv2017

Investigating the Magnetic Imprints of Major Solar Eruptions with SDO/HMI High-Cadence Vector Magnetograms

The solar active region photospheric magnetic field evolves rapidly during major eruptive events, suggesting appreciable feedback from the corona. Previous studies of these "magnetic imprints" are mostly based on line-of-sight only or lower-cadence vector observations; a temporally resolved depiction of the vector field evolution is hitherto lacking. Here, we introduce the high-cadence (90~s or 135~s) vector magnetogram dataset from the Helioseismic and Magnetic Imager (HMI) that is well suited for investigating the phenomenon. These observations allow quantitative characterization of the permanent, step-like changes that are most pronounced in the horizontal field component ($B_h$). A highly structured pattern emerges from analysis of an archetypical event, \texttt{SOL2011-02-15T01:56}, where $B_h$ near the main polarity inversion line increases significantly during the earlier phase of the associated flare with a time scale of several minutes, while $B_h$ in the periphery decreases at later times with smaller magnitudes and a slightly longer time scale. The dataset also allows effective identification of the "magnetic transient" artifact, where enhanced flare emission alters the Stokes profiles and the inferred magnetic field becomes unreliable. Our results provide insights on the momentum processes in solar eruptions. The dataset may also be useful to the study of sunquakes and data-driven modeling of the corona.

astro-ph.SR↗
arXiv2017

Face Detection using Deep Learning: An Improved Faster RCNN Approach

In this report, we present a new face detection scheme using deep learning and achieve the state-of-the-art detection performance on the well-known FDDB face detetion benchmark evaluation. In particular, we improve the state-of-the-art faster RCNN framework by combining a number of strategies, including feature concatenation, hard negative mining, multi-scale training, model pretraining, and proper calibration of key parameters. As a consequence, the proposed scheme obtained the state-of-the-art face detection performance, making it the best model in terms of ROC curves among all the published methods on the FDDB benchmark.

cs.CV↗
arXiv2016

Evolution of Magnetic Field and Energy in A Major Eruptive Active Region Based on SDO/HMI Observation

We report the evolution of magnetic field and its energy in NOAA active region 11158 over 5 days based on a vector magnetogram series from the Helioseismic and Magnetic Imager (HMI) on board the Solar Dynamic Observatory (SDO). Fast flux emergence and strong shearing motion led to a quadrupolar sunspot complex that produced several major eruptions, including the first X-class flare of Solar Cycle 24. Extrapolated non-linear force-free coronal fields show substantial electric current and free energy increase during early flux emergence near a low-lying sigmoidal filament with sheared kilogauss field in the filament channel. The computed magnetic free energy reaches a maximum of ~2.6e32 erg, about 50% of which is stored below 6 Mm. It decreases by ~0.3e32 erg within 1 hour of the X-class flare, which is likely an underestimation of the actual energy loss. During the flare, the photospheric field changed rapidly: horizontal field was enhanced by 28% in the core region, becoming more inclined and more parallel to the polarity inversion line. Such change is consistent with the conjectured coronal field "implosion", and is supported by the coronal loop retraction observed by the Atmospheric Imaging Assembly (AIA). The extrapolated field becomes more "compact" after the flare, with shorter loops in the core region, probably because of reconnection. The coronal field becomes slightly more sheared in the lowest layer, relaxes faster with height, and is overall less energetic.

astro-ph.SR↗
arXiv2016

Comparison of Coronal Extrapolation Methods for Cycle 24 Using HMI Data

Two extrapolation models of the solar coronal magnetic field are compared using magnetogram data from the SDO/HMI instrument. The two models, a horizontal current-current sheet-source surface (HCCSSS) model and a potential field-source surface (PFSS) model differ in their treatment of coronal currents. Each model has its own critical variable, respectively the radius of a cusp surface and a source surface, and it is found that adjusting these heights over the period studied allows better fit between the models and the solar open flux at 1 AU as calculated from the Interplanetary Magnetic Field (IMF). The HCCSSS model provides the better fit for the overall period from 2010 November to 2015 May as well as for two subsets of the period - the minimum/rising part of the solar cycle, and the recently-identified peak in the IMF from mid-2014 to mid-2015 just after solar maximum. It is found that a HCCSSS cusp surface height of 1.7 Rsun provides the best fit to the IMF for the overall period, while 1.7 & 1.9 Rsun give the best fits for the two subsets. The corresponding values for the PFSS source surface height are 2.1, 2.2 and 2.0 Rsun respectively. This means that the HCCSSS cusp surface rises as the solar cycle progresses while the PFSS source surface falls.

astro-ph.SR↗
arXiv2015

Evidence of thermal conduction suppression in hot coronal loops: Supplementary results

Slow magnetoacoustic waves were first detected in hot ($>$6 MK) flare loops by the SOHO/SUMER spectrometer as Doppler shift oscillations in Fe XIX and Fe XXI lines. Recently, such longitudinal waves have been found by SDO/AIA in the 94 and 131 Å channels. Wang et al. (2015) reported the first AIA event revealing signatures in agreement with a fundamental standing slow-mode wave, and found quantitative evidence for thermal conduction suppression from the temperature and density perturbations in the hot loop plasma of $\gtrsim$ 9 MK. The present study extends the work of Wang et al. (2015) by using an alternative approach. We determine the polytropic index directly based on the polytropic assumption instead of invoking the linear approximation. The same results are obtained as in the linear approximation, indicating that the nonlinearity effect is negligible. We find that the flare loop cools slower (by a factor of 2-4) than expected from the classical Spitzer conductive cooling, approximately consistent with the result of conduction suppression obtained from the wave analysis. The modified Spitzer cooling timescales based on the nonlocal conduction approximation are consistent with the observed, suggesting that nonlocal conduction may account for the observed conduction suppression in this event. In addition, the conduction suppression mechanism predicts that larger flares may tend to be hotter than expected by the EM-$T$ relation derived by Shibata & Yokoyama (2002)

astro-ph.SR↗
arXiv2015

Complex Flare Dynamics Initiated by a Filament-Filament Interaction

We report on an eruption involving a relatively rare filament-filament interaction on 2013 June 21, observed by SDO and STEREO-B. The two filaments were separated in height with a 'double-decker' configuration. The eruption of the lower filament began simultaneously with a descent of the upper filament resulting in a convergence and direct interaction of the two filaments. The interaction was accompanied by the heating of surrounding plasma and an apparent crossing of a loop-like structure through the upper filament. The subsequent coalescence of the filaments drove a bright front ahead of the erupting structures. The whole process was associated with a C3.0 flare followed immediately by an M2.9 flare. Shrinking loops and descending dark voids were observed during the M2.9 flare at different locations above a C-shaped flare arcade as part of the energy release, giving us a unique insight into the flare dynamics.

astro-ph.SR↗
arXiv2015

Evidence of thermal conduction suppression in a solar flaring loop by coronal seismology of slow-mode waves

Analysis of a longitudinal wave event observed by the Atmospheric Imaging Assembly (AIA) onboard the Solar Dynamics Observatory (SDO) is presented. A time sequence of 131 A images reveals that a C-class flare occurred at one footpoint of a large loop and triggered an intensity disturbance (enhancement) propagating along it. The spatial features and temporal evolution suggest that a fundamental standing slow-mode wave could be set up quickly after meeting of two initial disturbances from the opposite footpoints. The oscillations have a period of ~12 min and a decay time of ~9 min. The measured phase speed of 500$\pm$50 km/s matches the sound speed in the heated loop of ~10 MK, confirming that the observed waves are of slow mode. We derive the time-dependent temperature and electron density wave signals from six AIA extreme-ultraviolet (EUV) channels, and find that they are nearly in phase.The measured polytropic index from the temperature and density perturbations is 1.64$\pm$0.08 close to the adiabatic index of 5/3 for an ideal monatomic gas. The interpretation based on a 1D linear MHD model suggests that the thermal conductivity is suppressed by at least a factor of 3 in the hot flare loop at 9 MK and above. The viscosity coefficient is determined by coronal seismology from the observed wave when only considering the compressive viscosity dissipation. We find that to interpret the rapid wave damping, the classical compressive viscosity coefficient needs to be enhanced by a factor of 15 as the upper limit.

astro-ph.SR↗
arXiv2015

The Role of Erupting Sigmoid in Triggering a Flare with Parallel and Large-Scale Quasi-Circular Ribbons

In this paper, we present observations and analysis of an interesting sigmoid formation, eruption and the associated flare that occurred on 2014 April 18 using multi-wavelength data sets. We discuss the possible role of the sigmoid eruption in triggering the flare, which consists of two different set of ribbons: parallel ribbons as well as a large-scale quasi-circular ribbon. Several observational evidence and nonlinear force-free field extrapolation results show the existence of a large-scale fan-spine type magnetic configuration with a sigmoid lying under a section of the fan dome. The event can be explained with the following two phases. During the pre-flare phase, we observed the formation and appearance of sigmoid via tether-cutting reconnection between the two sets of sheared fields under the fan dome. The second, main flare phase, features the eruption of the sigmoid, the subsequent flare with parallel ribbons, and a quasi-circular ribbon. We propose the following multi-stage successive reconnections scenario for the main flare. First, tether-cutting reconnection is responsible for the formation and the eruption of the sigmoid structure. Second, the reconnection occurred in the wake of the erupting sigmoid produces the parallel flare ribbons on the both sides of the circular polarity inversion line. Third, the null-type reconnection higher in the corona, possibly triggered by the erupting sigmoid, leads to the formation of a large quasi-circular ribbon. For the first time we suggest a mechanism for this type of flare consisting of a double set of ribbons triggered by an erupting sigmoid in a large scale fan-spine type magnetic configuration.

astro-ph.SR↗
arXiv2015

Photospheric Electric Fields and Energy Fluxes in the Eruptive Active Region NOAA 11158

How much electromagnetic energy crosses the photosphere in evolving solar active regions? With the advent of high-cadence vector magnetic field observations, addressing this fundamental question has become tractable. In this paper, we apply the "PTD-Doppler-FLCT-Ideal" (PDFI) electric field inversion technique of Kazachenko et al. (2014) to a 6-day HMI/SDO vector magnetogram and Doppler velocity sequence, to find the electric field and Poynting flux evolution in active region NOAA 11158, which produced an X2.2 flare early on 2011 February 15. We find photospheric electric fields ranging up to $2$ V/cm. The Poynting fluxes range from $[-0.6$ to $2.3]\times10^{10}$ ergs$\cdot$cm$^{-2}$s$^{-1}$, mostly positive, with the largest contribution to the energy budget in the range of $[10^9$-$10^{10}]$ ergs$\cdot$cm$^{-2}$s$^{-1}$. Integrating the instantaneous energy flux over space and time, we find that the total magnetic energy accumulated above the photosphere from the initial emergence to the moment before the X2.2 flare to be $E=10.6\times10^{32}$ ergs, which is partitioned as $2.0$ and $8.6\times10^{32}$ ergs, respectively, between free and potential energies. Those estimates are consistent with estimates from preflare non-linear force-free field (NLFFF) extrapolations and the Minimum Current Corona estimates (MCC), in spite of our very different approach. This study of photospheric electric fields demonstrates the potential of the PDFI approach for estimating Poynting fluxes and opens the door to more quantitative studies of the solar photosphere and more realistic data-driven simulations of coronal magnetic field evolution.

astro-ph.SR↗