We present a catalog of 948,216 stars with mass labels and a catalog of 163,105 red clump (RC) stars with mass and age labels simultaneously. The training data set is crossmatched from the Large Sky Area Multi-Object Fiber Spectroscopic Telescope DR5, and high-resolution asteroseismology data, mass, and age are predicted by the random forest (RF) method or a convex-hull algorithm. The stellar parameters with a high correlation with mass and age are extracted and the test data set shows that the median relative error of the prediction model for the mass of the large sample is 3%, and for the mass and age of RC stars is 4% and 7%. We also compare the predicted age of RC stars with recent works and find that the final uncertainty of the RC sample could reach 18% for age and 9% for mass; meanwhile, the final precision of the mass for the large sample with different types of stars could reach 13% without considering systematics. All of this implies that this method could be widely used in the future. Moreover, we explore the performance of different machine-learning methods for our sample, including Bayesian linear regression and the gradient-boosting decision tree (GBDT), multilayer perceptron, multiple linear regression, RF, and support vector regression methods. Finally, we find that the performance of a nonlinear model is generally better than that of a linear model, and the GBDT and RF methods are relatively better.

We present a catalog of 948,216 stars with mass label and a catalog of 163,105 red clump (RC) stars with mass and age labels simultaneously. The training dataset is cross matched from the LAMOST DR5 and high resolution asteroseismology data, mass and age are predicted by random forest method or convex hull algorithm. The stellar parameters with high correlation with mass and age are extracted and the test dataset shows that the median relative error of the prediction model for the mass of large sample is 0.03 and meanwhile, the mass and age of red clump stars are 0.04 and 0.07. We also compare the predicted age of red clump stars with the recent works and find that the final uncertainty of the RC sample could reach 18% for age and 9% for mass, in the meantime, final precision of the mass for large sample with different type of stars could reach 13% without considering systematics, all these are implying that this method could be widely used in the future. Moreover, we explore the performance of different machine learning methods for our sample, including bayesian linear regression (BYS), gradient boosting decision Tree (GBDT), multilayer perceptron (MLP), multiple linear regression (MLR), random forest (RF) and support vector regression (SVR). Finally we find that the performance of nonlinear model is generally better than that of linear model, and the GBDT and RF methods are relatively better.

Primal-Dual scheme is particularly suitable for solving the non-smooth Total Variation (TV) model in imaging, and the soft thresholding algorithm is simple and effective for the Curvelet prior. We propose a hybrid prior of TV and Curvelet Prior (TVCP) model for the image restoration problems. In order to obtain high restoration quality, we propose Primal-Dual and Soft Threshold (PDST) algorithm to solve this convex optimization model (TVCP). Our inpainting experimental results have shown that PDST algorithm significantly outperforms PrimalDual for TV (PDTV) and Primal-Dual for Curvelet (PDC), in both subjective and objective image quality. Furthermore, TVCP model and PDST algorithm can be easily applied to solving other challenging problems in image, such as denosing, deconvolution, compressed sensing etc.