知识卡片
反向推断网络不是反向因果边
额外训练q(K|P)可以帮助在观测P后推断K,充当推断提议分布。它不是把DGP中的K→P改成P→K;生成图与推断图可有不同方向,应分别标明用途。
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
- · 因果与大语言模型
原书出处 · 1 处
- · 查看原书依据
Finally, we’ll train another model for inferring the King vignette given a Prince vignette. We’ll use this in inference later. Listing 13.13 Function to train the Prince to King model p2k_data = create_seq2seq_datasets( df["Prince"], df["King"]) train_dataset_prince2king, val_dataset_prince2king = p2k_data prince2king_model = train_seq2seq_model( prince2king_model_path, train_dataset_prince2king, val_dataset_prince2king, epochs=6 ) Running the preceding training procedure will take some time, especially if you’re not using GPU. Fortunately, there are
参考来源
- 位置:《Causal AI》第13章《因果与大语言模型》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Causal AI/_epub-src/OEBPS/Text/chapter-13.xhtml)
- 结论依据:摘录中关于反向推断网络不是反向因果边的定义、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:Finally, we’ll train another model for inferring the King vignette given a Prince vignette. We’ll use this in inference later. Listing 13.13 Function to train the Prince to King model p2k_data = create_seq2seq_datasets( df["Prince"], df["King"]) train_dataset_prince2king, val_dataset_prince2king = p2k_data prince2king_model = train_seq2seq_model( prince2king_model_path, train_dataset_prince2king, val_dataset_prince2king, epochs=6 ) Running the preceding training procedure will take some time, especially if you’re not using GPU. Fortunately, there are