Automatic Determination of Stellar Atmospheric Parameters Using Neural Networks and Instance-Based Machine Learning

In this article we show how machine learning methods can beeffectively applied to the problem of automatically predictingstellar atmospheric parameters from spectral information, a veryimportant problem in stellar astronomy. We apply feedforwardneural networks, Kohonen's self-organizing maps andlocally-weighted regression to predict the stellar atmosphericparameters effective temperature, surface gravity and metallicityfrom spectral indices. Our experimental results show that thethree methods are capable of predicting the parameters with verygood accuracy. Locally weighted regression gives slightly betterresults than the other methods using the original dataset asinput, while self-organizing maps outperform the other methods when significant amounts of noise are added. We also implemented a heterogeneous ensemble of predictors, combining the results given by the three algorithms. This ensemble yields better results than any of the three algorithms alone, using both the original and the noisy data.     

Short-Term Solar Flare Prediction Using a Sequential Supervised Learning Method

Solar flares are powered by the energy stored in magnetic fields, so evolutionary information of the magnetic field is important for short-term prediction of solar flares. However, the existing solar flare prediction models only use the current information of the active region. A sequential supervised learning method is introduced to add the evolutionary information of the active region into a prediction model. The maximum horizontal gradient, the length of the neutral line, and the number of singular points extracted from SOHO/MDI longitudinal magnetograms are used in the model to describe the nonpotentiality and complexity of the photospheric magnetic field. The evolutionary characteristics of the predictors are analyzed by using autocorrelation functions and mutual information functions. The analysis results indicate that a flare is influenced by the 3-day photospheric magnetic field information before flare eruption. A sliding-window method is used to add evolutionary information of the predictors into machine learning algorithms, then C4.5 decision tree and learning vector quantization are employed to predict the flare level within 48 hours. Experimental results indicate that the performance of the short-term solar flare prediction model within the sequential supervised learning framework is significantly improved.     

基于深度学习的LAMOST星系星族参数测量

星系的光谱包含其内部恒星的年龄和金属丰度等信息, 从观测光谱数据中测量这些信息对于深入了解星系的形成和演化至关重要. LAMOST (Large Sky Area Multi-Object Fiber Spectroscopic Telescope)巡天发布了大量的星系光谱, 这些高维光谱与它们的物理参数之间存在着高度的非线性关系. 而深度学习适合于处理多维、海量的非线性数据, 因此基于深度学习技术构建了一个8个卷积层$+$4个池化层$+$1个全连接层的卷积神经网络, 对LAMOST Data Release 7 (DR7)星系的年龄和金属丰度进行自动估计. 实验结果表明, 使用卷积神经网络通过星系光谱预测的星族参数与传统方法基本一致, 误差在0.18dex以内, 并且随着光谱信噪比的增大, 预测误差越来越小. 实验还对比了卷积神经网络与随机森林回归模型、深度神经网络的参数测量结果, 结果表明卷积神经网络的结果优于其他两种回归模型.     

Prediction of Short Stellar Activity Cycles using Derived and Established Empirical Relations between Activity and Rotation Periods

In our previous work,we investigated the occurrence rate of super-flares on various types of stars and their statistical properties,with a particular focus on G-type dwarfs,using entire Kepler data.The said study also considered how the statistics change with stellar rotation period,which in turn,had to be determined.Using such new data,as a by-product,we found 138 Kepler IDs of F-and G-type main sequence stars with rotation periods less than a day(P_rot <1 day).On one hand,previous...     

深度学习在天文学中的应用与改进

近年来,深度学习和人工智能技术迅猛发展,在多个学科领域得到了广泛关注和应用。天文学研究也不甘落后,涌现出一大批应用深度学习进行数据分析的工作。总结了深度学习在天文中的应用情况和趋势、天文数据类型和机器学习任务、天文中常用的深度学习网络模型和方法,以及深度学习在天文研究中的代表性应用和进展,并探讨和提出了其未来在天文学领域中的应用和改进建议。     

国内外空间天气短期预报和警报及其现状

综述了空间天气短期预报和警报所研究的内容和与之相对应的研究方法。同时，介绍了目前国内外在空间天气短期预报和警报方面的研究现状，及目前国内外从事这方面研究的主要机构或组织的概况。     

基于卷积神经网络的恒星光谱型和光度型的分类模型

恒星光谱分类是天文学中一个重要的研究问题.对于已经采集到的海量高维恒星光谱数据的分类,采用模式匹配方法对光谱型分类较为成功,但其缺点在于标准恒星模版之间的差异性在匹配实际观测数据中不能体现出来,尤其是当需要进行光谱型和光度型的二元分类时模版匹配法往往会失败.而采用谱线特征测量的光度型分类强烈地依赖谱线拟合的准确性.为了解决二元分类的问题,介绍了一种基于卷积神经网络的恒星光谱型和光度型分类模型(Classification model of Stellar Spectral type and Luminosity type based on Convolution Neural Network, CSSL CNN).这一模型使用卷积神经网络来提取光谱的特征,通过注意力模块学习到了重要的光谱特征,借助池化操作降低了光谱的维度并压缩了模型参数的数量,使用全连接层来学习特征并对恒星光谱进行分类.实验中使用了大天区面积多目标光纤光谱天文望远镜(Large Sky Area Multi-Object Fiber Spectroscopy Telescope, LAMOST)公开数据集Data Release 5 (DR5,用了其中71282条恒星光谱数据,每条光谱包含了3000多维的特征)对该模型的性能进行验证与评估.实验结果表明,基于卷积神经网络的模型在恒星的光谱型分类上准确率达到92.04%,而基于深度神经网络的模型(Celestial bodies Spectral Classification Model, CSC Model)只有87.54%的准确率; CSSL CNN在恒星的光谱型和光度型二元分类上准确率达到83.91%,而模式匹配方法MKCLASS仅有38.38%的准确率且效率较低.     

空间碎片天基光电光学可见条件与预报

随着航天活动的增加，产生的空间碎片也越来越多，空间环境日趋恶化，已经对人类的空间活动构成了威胁。监视测量这些空间碎片，天基光电比地基光电更为有利，而天基光电的光学可见条件与地基光电相比，有相似的，也有不同的，针对天基光电，给出了空间碎片的光学可见条件，即日光条件、地影条件、地光条件、地球背景条件、月光条件。在天基光电轨道特征、光学可见条件及天基光电坐标系已知的情况下，建立起天基光电预报方法。既可用于空间碎片预报，也可以用于空间碎片的轨道识别。     

Comparison of BDS Station Clock Short-term Prediction Models and their Applications in Precise Orbit Determination

BDS (BeiDou Navigation Satellite System) ground tracking stations are equipped with high accuracy atomic clocks, and they are synchronized with the BDS time scale (BDT) via the Precise Orbit Determination (POD) processing. During the periods of satellite maneuver and post-maneuver, station clocks are kept fixed as known values in the POD processing. To improve the real-time POD capability, station clocks need to be predicted. In this paper, the performance of three clock prediction models is evaluated, including quadratic polynomial model (QP), periodical term model (PM), and grey model (GM). The precision of clock fitting and prediction, as well as the performance of the prediction models in POD are compared. Data of six stations are used for test, and the results show that: the mean fitting accuracy of quadratic polynomial model, periodical term model, and grey model is 0.14 ns, 0.05 ns, 0.27 ns, respectively; the 1 h and 2 h prediction precision of the three models is 1.17 ns, 0.88 ns, 1.28 ns, and 2.72 ns, 2.09 ns, 2.53 ns, respectively. Applying the 1 h and 2 h predicted station clocks in the POD, the 3D orbit accuracy reaches the best using the periodical term model, while the radial accuracy of satellite orbit is rather close for the three models with the difference within 3 cm.     

北斗系统测站钟差短期预报模型比较及其在单星定轨中的应用

北斗卫星导航系统(BDS)地面跟踪站都配置有高精度的氢原子钟,并基于精密定轨数据处理与主站的时间基准进行同步.在卫星轨道机动以及机动恢复期间,通常采用几何法定轨以及单星定轨确定卫星的轨道.而在这两种定轨模式中,需要提供精确的测站钟差作为输入.为提高定轨的实时性,需要对测站钟差进行预报处理.分析了2次多项式模型、附加周期项模型、灰色模型3种模型对北斗地面跟踪站钟差短期拟合和预报的性能,并将钟差预报结果应用于单星定轨,同时还分析了不同预报钟差用于定轨的精度.试验发现,以上3种模型对6个测站钟差的平均拟合精度分别为0.14 ns、0.05 ns、0.27 ns,预报1 h的平均精度分别为1.17 ns、0.88 ns、1.28 ns,预报2 h的平均精度分别为2.72 ns、2.09 ns、2.53 ns.采用3种模型对测站钟差进行预报并用于单星定轨,采用附加周期项的钟差预报模型轨道3维误差最小,不同模型轨道径向精度差异在3 cm以内.以上结果表明,附加周期项的站钟拟合及预报模型在北斗系统机动期间的轨道恢复数据处理具有最好的效果.     

倾角函数展开及其在分析法轨道预报中的应用

田谐项摄动是分析法轨道预报中的重要部分,其中包含大量倾角函数及其偏导数的计算.由于具有精度更高、速度更快的优点,倾角函数一般通过递推方法计算.以文献中提出的改进Gooding方法为基础,将其给出的程序稍加改进,在计算2–50阶倾角函数时缩短了约24%的计算时间.考虑到分析法预报过程中轨道平倾角变化很小,以泰勒展开式计算倾角函数,可极大提高计算速度,较大程度地减小分析法预报耗时,且引力场阶次越高,减小幅度越大,取50阶时预报耗时缩短了48%.另一方面,以2阶展开式计算倾角函数时,与改进Gooding法相比,分析法预报星历偏差很小.对于500 km高度的低轨卫星,分别以改进Gooding法和2阶泰勒展开式计算倾角函数,预报3天,当地球引力场阶次不高于50时,二者预报星历偏差RMS (Root Mean Square)低于1 mm,且随着轨道高度的增加,预报星历偏差RMS逐渐减小.     

Prediction of satellite orbits contraction due to diurnally varying oblate atmosphere and altitude-dependent scale height using KS canonical elements

A new non-singular analytical theory for the motion of near-Earth satellite orbits with the air drag effect is developed in terms of uniformly regular KS canonical elements. Diurnally varying oblate atmosphere is considered with variation in density scale height dependent on altitude. The series expansion method is utilized to generate the analytical solutions and terms up to fourth-order terms in eccentricity and c (a small parameter dependent on the flattening of the atmosphere) are retained. Only two of the nine equations are solved analytically to compute the state vector and change in energy at the end of each revolution, due to symmetry in the equations of motion. The important drag perturbed orbital parameters: semi-major axis and eccentricity are obtained up to 500 revolutions, with the present analytical theory and by numerical integration over a wide range of perigee height, eccentricity and inclination. The differences between the two are found to be very less. A comparison between the theories generated with terms up to third- and fourth-order terms in c and e shows an improvement in the computation of the orbital parameters semi-major axis and eccentricity, up to 9%. The theory can be effectively used for the re-entry of the near-Earth objects, which mainly decay due to atmospheric drag.