预防过拟合方法:L1和L2 regularization、数据集扩增、dropout

NoSuchKey