知识卡片

普通读书笔记卡 · 6723 · 生成模型:GPT-6

丢弃层作用于训练中的激活

理解等级:中级

dropout训练时随机将部分激活置零,减少对固定通道组合的依赖,推断时通常关闭并按实现缩放。它是正则方案,不保证对抗鲁棒或任何网络都受益;放哪里及概率大小需验证,p=0实际上无随机丢弃。

这张卡片看不懂?

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

      One popular technique for increasing the robustness of neural networks is random dropout. Random dropout, or just dropout, has become popular because of its ease and effectiveness. Your neural networks will almost always benefit from a dropout layer. A dropout layer randomly hides some of the neuron outputs from the neurons listening to them. This causes that pathway in your artificial brain to go quiet

    参考来源

    - 位置:《Natural Language Processing in Action, Second Edition》第07章《卷积网络识别文本模式》(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-7.html) - 结论依据:摘录中关于丢弃层作用于训练中的激活的定义、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:One popular technique for increasing the robustness of neural networks is random dropout. Random dropout, or just dropout, has become popular because of its ease and effectiveness. Your neural networks will almost always benefit from a dropout layer. A dropout layer randomly hides some of the neuron outputs from the neurons listening to them. This causes that pathway in your artificial brain to go quiet