知识卡片
循环层丢弃设置要看层数与位置
层间dropout可减少部分共适应,具体库通常只在多层循环层之间应用,单层配置未必起效。增加丢弃不保证修好任何过拟合,20%到50%也不是普遍安全区间;应记录实际生效位置与验证表现。
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
原书出处 · 1 处
- · 查看原书依据
If you feel the model is overfitting the training data, but you can’t find a way to make your model simpler, you can always try increasing the Dropout(percentage). This is like a sledgehammer that reduces overfitting while allowing your model to have as much complexity as it needs to match the data. If you set the dropout percentage much higher than 50%, the model starts
参考来源
- 位置:《Natural Language Processing in Action, Second Edition》第08章《循环网络与长短期记忆》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Natural Language Processing in Action, Second Edition/_epub-src/OEBPS/Text/chapter-8.html)
- 结论依据:摘录中关于循环层丢弃设置要看层数与位置的定义、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:If you feel the model is overfitting the training data, but you can’t find a way to make your model simpler, you can always try increasing the Dropout(percentage). This is like a sledgehammer that reduces overfitting while allowing your model to have as much complexity as it needs to match the data. If you set the dropout percentage much higher than 50%, the model starts