知识卡片
证据下界由后验差距决定
对固定观测,有log pθ(x)=ELBO(x)+KL(qψ(z|x)||pθ(z|x)),所以ELBO不大于对数似然。下界等号要求近似后验与模型后验一致,模型容量、变分族和优化都可能留下差距。单独提高下界不等于精确最大化真实似然。
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
The VAE objective, J, is often called the Evidence Lower Bound or ELBO, because it is a lower-bound on the data log-likelihood (i.e. Jp, which equals log pθ(x)). log pθ(x) is sometimes referred to as the evidence for x. Using the definition of KL-divergence, we can also rewrite J as follows: (33.33) In this form, the VAE objective is presented as the sum of two
参考来源
- 位置:《Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)》第33章《生成建模与表示学习》(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/c033.xhtml)
- 结论依据:摘录中关于证据下界由后验差距决定的定义、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 VAE objective, J, is often called the Evidence Lower Bound or ELBO, because it is a lower-bound on the data log-likelihood (i.e. Jp, which equals log pθ(x)). log pθ(x) is sometimes referred to as the evidence for x. Using the definition of KL-divergence, we can also rewrite J as follows: (33.33) In this form, the VAE objective is presented as the sum of two