5,337
edits
Deep Learning: Difference between revisions
→Bias-variance tradeoff
| Line 402: | Line 402: | ||
<math>G \approx E_{\sigma}[G'] = \frac{2}{n} E_{\sigma} \left[ \sup_{h \in H} \sum_{i=1}^{n} \sigma_i f(z^{(i)}) \right]</math> | <math>G \approx E_{\sigma}[G'] = \frac{2}{n} E_{\sigma} \left[ \sup_{h \in H} \sum_{i=1}^{n} \sigma_i f(z^{(i)}) \right]</math> | ||
<math>R(A) = \frac{1}{n} E_{\sigma} \left[ \sup _a \in A \sum_{i=1}^{n} \sigma_i a_i \right]</math> | <math>R(A) = \frac{1}{n} E_{\sigma} \left[ \sup _a \in A \sum_{i=1}^{n} \sigma_i a_i \right]</math> | ||
is called the Rademacher complexity. | |||
Setting <math>A = \{a_i\} = F \circ S</math> | |||
<math>\implies G \approx 2 R(F \circ S)</math> | <math>\implies G \approx 2 R(F \circ S)</math> | ||