知识卡片
归纳偏置是学习偏好的来源
架构、目标函数和先验会让学习算法偏好某些解释。因果图及机制独立假设也是归纳偏置,可在适用时限制模型空间。偏置的价值取决于是否符合问题,而非只要带有因果名称就优于其他约束。
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
assumptions (explicit or implicit) that lead an inference algorithm to prefer certain inferences or predictions over others. Examples of inductive bias include Occam’s Razor and the assumption in forecasting that trends in the past will continue into the future.
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:摘录中关于归纳偏置是学习偏好的来源的定义、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:assumptions (explicit or implicit) that lead an inference algorithm to prefer certain inferences or predictions over others. Examples of inductive bias include Occam’s Razor and the assumption in forecasting that trends in the past will continue into the future.