中文说明：贝叶斯网络结构学习的K2算法。在确定有节点序的情况下，依据K2评分函数，从空图开始加边，每加一条边，计算当前网络的K2评分，若评分高于之前的，则加上这一条边，否则就不加。因为这是在节点序已定的情况下学的，不需要转边。然后再执行减边操作，一次减边，若得分增高，则删去当前边，否则不删。
English Description：
K2 algorithm of Bayesian network structure learning. In the case of determining the node order, according to the K2 scoring function, add edges from the empty graph. Each edge is added to calculate the K2 score of the current network. If the score is higher than the previous one, this edge will be added, otherwise it will not be added. Because this is learned when the node order has been determined, there is no need to turn the edge. Then perform the edge subtraction operation again. If the score increases, the current edge will be deleted, otherwise it will not be deleted.