分位数回归模型自提出以来，已逐步成为统计与经济学中的一个重要分析方法，由于其对比于传统的最小二乘回归能更细致地刻画分布的特征，以及具有更稳健的算法，使之广泛应用于经济、医学、环境科学、生存分析和动植物学等方面。另一方面，minimum error entropy算法则是根据信息论中的 Renyi熵和Shannon熵构造的一种最小化熵的算法。 该算法最小化了误差信号的熵，从而使得所获得的信息最大化，被用于自适应信号系统处理的各个方面。.而理论分析上，Steinwart等在2008及2011年的文章中首先对分位数回归做出了分析，胡婷等2013年文章中也首次对MEE算法给出了一个完整的收敛性分析。本项目基于这两个课题，拟在胡婷等人的工作基础上，将MEE算法中的误差函数换成pinball loss函数，试图进行误差分解，得到收敛性的结果和学习率。

The quantile regression model gradually becomes an important method in statistics and economics after proposed. Due to its preciseness in characterizing distribution of variables and robustness in algorithm comparing classical least square regression, it is widely used in economics, medical science, environmental science, survival analysis and thremmatology. On the other hand, minimum error entropy algorithm is an algorithm minimizing the entropy of error, which is constructed according to Renyi's entropy and Shannon's entropy in information theory. It minimizes the entropy of error signal, then the obtained information is maximized. So that it is applied in many branches of self adaptive signal process system..As for the theoretical analysis, Steinwart etc. firstly studied quantile regression in 2008 and 2011, while Ting Hu etc. conducted a complete convergence analysis in 2013 for MEE algorithm for the first time. Under the framework of these two relatively novel topics, we aim to follow the way of Ting Hu etc.'s paper in this project. After giving an error decomposition to MEE algorithm with pinball loss function, we expect to have a convergence result and learning rate.