说明：  早期的基于样本学习的字典学习算法，一般MOD算法需要几十次迭代即可收敛是一个比较可行的方法。缺点在于运算中需要对矩阵求逆，造成计算量过大。
(In the early dictionary learning algorithm based on sample learning, the general mod algorithm needs dozens of iterations to converge. The disadvantage is that the matrix needs to be inversed in the operation, resulting in too much calculation.)