Reinforcement learning - Wikipedia
6
公开标注数
6
参与人数
2026-07-16 10:01:53
首次 Whisper
本页的公开 Whisper
划选高亮2026-07-16 16:25:53
原文高亮摘录
“The agent's action selection is modeled as a map called policy”
Whisper 随想笔记
So basically the policy is just a fancy probability table for actions.
划选高亮2026-07-16 13:22:53
原文高亮摘录
“The trade-off between exploration and exploitation has been most thoroughly studied through the multi-armed bandit problem”
Whisper 随想笔记
Burnetas and Katehakis, now that's a throwback to my grad school days.
划选高亮2026-07-16 13:13:53
原文高亮摘录
“The trade-off between exploration and exploitation has been most thoroughly studied through the multi-armed bandit problem”
Whisper 随想笔记
Classic bandit problem, but real life has way more than a few arms to pull.
划选高亮2026-07-16 10:19:53
原文高亮摘录
“reinforcement learning is particularly well-suited to problems that include a long-term versus short-term reward trade-off.”
Whisper 随想笔记
Makes me think of those delayed gratification experiments but for machines—so fitting.
划选高亮2026-07-16 10:10:53
原文高亮摘录
“reinforcement learning is particularly well-suited to problems that include a long-term versus short-term reward trade-off.”
Whisper 随想笔记
But isn't that true for any optimization problem? Kind of oversimplified.
划选高亮2026-07-16 10:01:53
原文高亮摘录
“reinforcement learning is particularly well-suited to problems that include a long-term versus short-term reward trade-off.”
Whisper 随想笔记
This is why RL always feels like the ultimate patience test—short-term pain for long-term gain.
分享本页 Whisper
短链接
https://domwhisper.com/s/144db3499e4f嵌入代码
<iframe src="https://domwhisper.com/embed/144db3499e4f" width="100%" height="480" style="border:0;border-radius:16px" loading="lazy"></iframe>