知识卡片
损失下降不证明推断已校准
变分损失趋稳只是优化诊断之一,不证明近似族足够、真实后验已恢复或反事实准确。应结合模型约束、证据回代、不同初始化与适当参考计算来评价;图中的训练曲线不能单独承担推断验收。
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
Figure 9.13 shows loss during training indicating variational inference converged. Figure 9.13 Loss during optimization of the parameters of the distribution approximating P(NSex, NFemur, NHeight|F=44, H=165) After training is completed, we extract the optimized parameters for our updated exogenous variable distribution. Listing 9.16 Extract parameters of updated exogenous distribution n_sex_p = param("p").item() #1 n_femur_loc = param("n_femur_loc").item() #1 n_femur_scale = param("n_femur_scale").item() #1 n_height_loc = param("n_height_loc").item() #1
参考来源
- 位置:《Causal AI》第09章《通用反事实推断算法》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Causal AI/_epub-src/OEBPS/Text/chapter-9.xhtml)
- 结论依据:摘录中关于损失下降不证明推断已校准的定义、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:Figure 9.13 shows loss during training indicating variational inference converged. Figure 9.13 Loss during optimization of the parameters of the distribution approximating P(NSex, NFemur, NHeight|F=44, H=165) After training is completed, we extract the optimized parameters for our updated exogenous variable distribution. Listing 9.16 Extract parameters of updated exogenous distribution n_sex_p = param("p").item() #1 n_femur_loc = param("n_femur_loc").item() #1 n_femur_scale = param("n_femur_scale").item() #1 n_height_loc = param("n_height_loc").item() #1