5,321
edits
Machine Learning: Difference between revisions
→Rademacher Complexity
| Line 185: | Line 185: | ||
====Definition==== | ====Definition==== | ||
The Rademacher complexity, like the VC dimension, measures how "rich" the hypothesis space is.<br> | The Rademacher complexity, like the VC dimension, measures how "rich" the hypothesis space is.<br> | ||
In this case, we see how well we can fit random noise.<br> | |||
Given a set <math>A \subset \mathbb{R}^m</math> the Rademacher complexity is:<br> | Given a set <math>A \subset \mathbb{R}^m</math> the Rademacher complexity is:<br> | ||
<math>R(A) = \frac{1}{m}E_{\sigma} [\sup_{a \in A} \sum_{i=1}^{m} \sigma_i a_i]</math><br> | <math>R(A) = \frac{1}{m}E_{\sigma} [\sup_{a \in A} \sum_{i=1}^{m} \sigma_i a_i]</math><br> | ||