Alex-McAvoy / Gitalk-Comment

存储博客评论的 Gittalk 仓库
0 stars 0 forks source link

VC 维与 Rademacher 复杂度 | Alex_McAvoy #1096

Open Alex-McAvoy opened 2 weeks ago

Alex-McAvoy commented 2 weeks ago

https://alex-mcavoy.github.io/artificial-intelligence/machine-learning/compututational-learning-theory/b3e4012d.html

【概述】在 PAC 学习理论概述 中,介绍了 PAC 学习理论,由于恰 PAC 学习并不实际,因此更重要的是研究假设空间 $\mathcal{H}$ 与概念类 $\mathcal{C}$ 不同的情景,即在给定 $n$ 个样本的训练集 $D$ 时,找出满足误差参数 $\epsilon$ 的假设 在 $|\mathcal{H}|$ 无限时,称假设空间 $\mathcal{H}$ 为无限假设空间,现实学