知识卡片
最大似然联系正向散度
KL(p数据||pθ)=E数据log p数据−E数据log pθ,第一项不依赖θ,因此在相应可积和支持条件下,最小化正向KL等价于最大化期望对数似然。实际以训练样本平均近似期望,有限样本最优不保证真实分布最优。
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
- · 生成模型
原书出处 · 1 处
- · 查看原书依据
The problem is that we do not actually have the function pdata, we only have samples from this function, x ∼ pdata. Therefore, we need a divergence D that measures the distance13 between pθ and pdata while only accessing samples from pdata. A common choice is to use the Kullback-Leibler (KL) divergence, which is defined as follows:
参考来源
- 位置:《Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)》第32章《生成模型》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)/_epub-src/OPS/xhtml/c032.xhtml)
- 结论依据:摘录中关于最大似然联系正向散度的定义、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:The problem is that we do not actually have the function pdata, we only have samples from this function, x ∼ pdata. Therefore, we need a divergence D that measures the distance13 between pθ and pdata while only accessing samples from pdata. A common choice is to use the Kullback-Leibler (KL) divergence, which is defined as follows: