← 返回刚才阅读的位置

知识卡片

普通读书笔记卡 · 6558.a · 生成模型:GPT-6

无监督解耦需要额外信息

理解等级:高级

仅凭无标签观测,潜在表示往往存在多种同样解释数据的坐标变换,无法保证恢复唯一真实因素。领域约束、干预或部分标签可增加信息,但仍需说明具体解决了哪种歧义;不宜把任意Z坐标操作直接称为真实干预。

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

      domain knowledge, interventions, and semi-supervision can help. See https://www.altdeep.ai/p/causalaibook for references on causal representation learning and disentanglement of causal factors. As we progress through the book, we’ll develop intuition for what the “causal information” in such a representation should look like.

    参考来源

    - 位置:《Causal AI》第05章《连接因果与深度学习》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Causal AI/_epub-src/OEBPS/Text/chapter-5.xhtml) - 结论依据:摘录中关于无监督解耦需要额外信息的定义、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:domain knowledge, interventions, and semi-supervision can help. See https://www.altdeep.ai/p/causalaibook for references on causal representation learning and disentanglement of causal factors. As we progress through the book, we’ll develop intuition for what the “causal information” in such a representation should look like.