知识卡片
欠拟合不能只靠增加层数解决
训练和验证都差时可能有容量不足,也可能优化失败、输入无信息或标签问题。先查学习过程与数据,再比较更合适表示和模型;简单将SVM换网络或不断加参数不保证修复。
这张卡片看不懂?
这张卡片已通过人工审查,不再接受失败反馈。
已有反馈 · 0 条
还没有人提交反馈。
在整书笔记中的位置
- · 训练机器学习模型
原书出处 · 1 处
- · 查看原书依据
Lastly, let us discuss two very common behaviors that all machine learning models are known to display: overfitting and underfitting. Both of these terms refer to the model not being able to properly generalize to unseen data and thus behave poorly during model evaluation. Overfitting in machine learning occurs when a model performs very well on a training set (e.g., very high accuracy and very
参考来源
- 位置:《Machine Learning Fundamentals: Concepts, Models, and Applications》第09章《训练机器学习模型》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Machine Learning Fundamentals: Concepts, Models, and Applications/_epub-src/OEBPS/xhtml/23_chapter09.xhtml)
- 结论依据:摘录中关于欠拟合不能只靠增加层数解决的定义、比较或条件支撑卡片;原书表述不严谨的范围已在覆盖记录逐项限定,不把案例结果外推为普遍结论。
- 原始内容:Lastly, let us discuss two very common behaviors that all machine learning models are known to display: overfitting and underfitting. Both of these terms refer to the model not being able to properly generalize to unseen data and thus behave poorly during model evaluation. Overfitting in machine learning occurs when a model performs very well on a training set (e.g., very high accuracy and very