知识卡片
有限状态不保证记住全部历史
循环模型可按同一规则继续处理长序列,但固定宽度和有限精度状态只保留任务学到的摘要。旧信息可能遗忘或干扰,训练也有长依赖困难;可继续读不等于无损记住无限文本,更不等于能还原原文。
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
8.1.2 RNNs remember everything you tell them Have you ever accidentally touched wet paint and found yourself “reusing” that paint whenever you touched something? Perhaps, as a child, you might have fancied yourself an impressionistic painter, as you shared your art with the world by finger-painting the walls around you. You’re about to learn how to build a more mindful impressionistic word painter. In chapter
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:摘录中关于有限状态不保证记住全部历史的定义、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:8.1.2 RNNs remember everything you tell them Have you ever accidentally touched wet paint and found yourself “reusing” that paint whenever you touched something? Perhaps, as a child, you might have fancied yourself an impressionistic painter, as you shared your art with the world by finger-painting the walls around you. You’re about to learn how to build a more mindful impressionistic word painter. In chapter