A machine learning technique with two-dimension convolutional neural network is proposed for detecting exoplanet transits. To test this new method, five different types of deep learning models with or without folding are constructed and studied. The light curves of the Kepler Data Release 25 are employed as the input of these models. The accuracy, reliability, and completeness are determined and their performances are compared. These results indicate that a combination of two-dimension convolutional neural network with folding would be an excellent choice for the future transit analysis.

The frequency ratios and of HD 49933 exhibit an increase at high frequencies. This behavior also exists in the ratios of other stars, which is considered to result from the low signal-to-noise ratio and the larger line width at the high-frequency end and could not be predicted by stellar models in previous work. Our calculations show that the behavior not only can be reproduced by stellar models, but can be predicted by asymptotic formulas of the ratios. The frequency ratios of the Sun, too, can be reproduced well by the asymptotic formulas. The increased behavior derives from the fact that the gradient of mean molecular weight at the bottom of the radiative region hinders the propagation of p-modes, while the hindrance does not exist in the convective core. This behavior should exist in the ratios of stars with a large convective core. The characteristic of the ratios at high frequencies provides a strict constraint on stellar models and aids in determining the size of the convective core and the extent of overshooting. Observational constraints point to a star with $M=1.28\pm0.01 M_{\odot}$, $R=1.458\pm0.005 R_{\odot}$, $t=1.83\pm0.1$ Gyr, $r_{cc}=0.16\pm0.02 R_{\odot}$, $\alpha=1.85\pm0.05$, and $\delta_{ov}=0.6\pm0.2$ for HD 49933.