知识卡片
模拟监督携带生成先验
用许多模拟SCM生成数据并训练网络猜结构,训练集的结构频率和函数族构成归纳偏置。对观测无法区分的候选,网络偏好可来自模拟先验,而非目标数据新揭示了唯一真相。报告应区分数据证据与生成设置。
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
- · 结构因果模型
原书出处 · 1 处
- · 查看原书依据
Suppose then you fed the trained model data actual samples of three-sided die rolls, with the goal of predicting the ground-truth SCM. That predictive model’s prediction might favor a stick with the {1, 2, 3} ordering over the equivalent {2, 3, 1} ordering. But it would only do so if the {1, 2, 3} ordering was more common in the training data.
参考来源
- 位置:《Causal AI》第06章《结构因果模型》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Causal AI/_epub-src/OEBPS/Text/chapter-6.xhtml)
- 结论依据:摘录中关于模拟监督携带生成先验的定义、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:Suppose then you fed the trained model data actual samples of three-sided die rolls, with the goal of predicting the ground-truth SCM. That predictive model’s prediction might favor a stick with the {1, 2, 3} ordering over the equivalent {2, 3, 1} ordering. But it would only do so if the {1, 2, 3} ordering was more common in the training data.