依托中国虚拟天文台的互联网+天文大数据信息，从天文观测、数值模拟等途径获得完备的天文数据库，充分利用LAMOST特有的采集低分辨率光纤光谱大数据中能出现极低比例稀有天体的观测优势。本研究结合A型恒星天体特征值间的相关性分析，在LAMOST已发布近五十万条A型恒星光谱观测数据中学习AGB星和post-AGB星红外超巨星已有的成功研究发现，重点针对研究甚少且亮度非常高的蓝超巨星所包含尘埃、特征带及谱线特性，选取post-AGB星侯选小样本建立训练模型光谱，采用TensorFlow卷积神经网络架构高性能计算，融合WISE数据分析、证认、解释遥远的AGB星结构和演化模型，提供一套标准的发现、访问规则，整合新挖掘模式和特征相关性，进而发现新天体类型，探索恒星大气物理参数相关性自适应算法软件包，且具备可移植、易推广数据分析算法，融合相关性分析导出新模型，推动天文学研究发展进程。

Relying on the China Virtual Observatory of Internet and the big data astronomical information, which the astronomical observations, numerical simulation and other means to obtain the complete astronomical database.Taking full advantage of LAMOST unique low-resolution fiber-optic data acquisition can be observed in a very small proportion of rare earth observation advantages . In this study, based on the analysis of the correlation between the distance and the eigenvalue of type A stellar , we studied the existing super-stars of AGB and post-AGB in the nearly half a million A-type stellar spectra observed by LAMOST. The successful study found that focusing on the dust, characteristic bands and spectral characteristics of the blue super giant with little or very high brightness, A candidate sample of the post-AGB star is selected to establish the training model spectrum, TensorFlow convolutional neural network architecture is used for high-performance computing, combining WISE data to identify and analyze the remote AGB star structure and evolution model, and provides a set of standard discovery and access rules. , Integrate new mining patterns and feature correlations to discover new types of celestial bodies, explore stellar physics parametric adaptive adaptive software packages, and have portable, easy-to-promote data analysis algorithms, fusion correlation analysis to derive new models, and promote astronomy Study the development process.