知识卡片
过采样重复数据不增加独立信息
复制少数类提高其训练影响,不能创造新的动物或表达变式。大量重复可能让模型记住具体样例;应只在训练内进行,并用原目标分布及独立实体测试,不在切分前复制后分散到两边。
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
Oversampling is the technique of repeating examples from the underrepresented class or classes. Let’s take the dog–cat example from earlier (only 200 dogs to 20,000 cats). You can simply repeat the dog images you have 100 times each and end up with 40,000 total samples—half dogs and half cats.
参考来源
- 位置:《Natural Language Processing in Action, Second Edition》附录D《附录D:机器学习工具与技术》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Natural Language Processing in Action, Second Edition/_epub-src/OEBPS/Text/appendix-d.html)
- 结论依据:摘录中关于过采样重复数据不增加独立信息的定义、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:Oversampling is the technique of repeating examples from the underrepresented class or classes. Let’s take the dog–cat example from earlier (only 200 dogs to 20,000 cats). You can simply repeat the dog images you have 100 times each and end up with 40,000 total samples—half dogs and half cats.