知识卡片
即时反馈不等于轨迹独立
经典bandit假设通常限制环境的奖励机制,而学习策略会依赖历史行动和奖励,使实际行动序列相关。不能把每轮环境噪声独立,扩大成自适应收集的全部变量在时间上互相独立。
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
independent trials, meaning that the variables at the t th timestep are independent of variables at other timesteps. The term “bandit” comes from an analogy to “one-armed bandits,” which is a slang term for casino slot machines that traditionally have an arm that the player pulls to initiate gameplay. Slot machine gameplay provides bandit feedback—you deposit a token, pull the arm, and instantly find out
参考来源
- 位置:《Causal AI》第12章《因果决策与强化学习》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Causal AI/_epub-src/OEBPS/Text/chapter-12.xhtml)
- 结论依据:摘录中关于即时反馈不等于轨迹独立的定义、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:independent trials, meaning that the variables at the t th timestep are independent of variables at other timesteps. The term “bandit” comes from an analogy to “one-armed bandits,” which is a slang term for casino slot machines that traditionally have an arm that the player pulls to initiate gameplay. Slot machine gameplay provides bandit feedback—you deposit a token, pull the arm, and instantly find out