损失函数（Loss Function）是指用数学的方法衡量假设函数预测结果与真实值之间的误差。机器学习算法的任务就是要使这个损失函数的差距越来越小，达到越来越准确的目标。

A.The attribute values and classlabels may suffer from stochasticnoise，systematic noise，and random artefacts.
B.The quality of the induced classifier depends ontraining examples.
C.Apart from low error rate，some applications require that the classifer provides the reasons behind the classification.
D.The simplest performance criterion is rmse.