https://dextermayhewjd.github.io/tags/policy-gradients/Recent content in Policy Gradients on dextermayhewjdHugozh-cnWed, 07 Oct 2026 00:00:00 +0000https://dextermayhewjd.github.io/courses/cs285/lecture-05/Wed, 07 Oct 2026 00:00:00 +0000https://dextermayhewjd.github.io/courses/cs285/lecture-05/先逐步推导 REINFORCE，再解释 baseline、reward-to-go 与伪损失。每个知识节点旁都有公式，点击可进入详细拆解。https://dextermayhewjd.github.io/courses/cs285/lecture-09/Wed, 07 Oct 2026 00:00:00 +0000https://dextermayhewjd.github.io/courses/cs285/lecture-09/逐步拆开换分布、轨迹重要性权重、因果性、单步替代目标与 clip/min，保留公式逐行拆解和四篇卡点精读。