4.2 过、欠拟合(Underfitting and Overfitting)
4.2 过、欠拟合(Underfitting and Overfitting)
李沐
B站:https://space.bilibili.com/1567748478/channel/collectiondetail?sid=28144
课程主页:https://c.d2l.ai/stanford-cs329p/课件:https://c.d2l.ai/stanford-cs329p/_static/pdfs/cs329p_slides_6_2.pdf
1. 概念
训练误差:模型在训练集上的误差;
泛化误差:模型在新数据集上预测值和真实值的差(数据只用一次);
注:解释训练误差和泛化误差
2. 过拟合和欠拟合
-
训练误差低,泛化误