5,337
edits
Deep Learning: Difference between revisions
→Bias-variance tradeoff
| Line 414: | Line 414: | ||
Example: <math>H = \{ h(x) = w^t x, \Vert w \Vert_2 \leq 1\}</math> | Example: <math>H = \{ h(x) = w^t x, \Vert w \Vert_2 \leq 1\}</math> | ||
<math>R(H \circ S) \leq \frac{\max_i \Vert x^{(i)} \Vert_2}{\sqrt{n}}</math> | <math>R(H \circ S) \leq \frac{\max_i \Vert x^{(i)} \Vert_2}{\sqrt{n}}</math> | ||
Question: What is the Rademacher complexity of a deep model? | |||
<math>H = \{ h(x) \mid h \text{ is a NN with some structure}\}</math> | |||
If <math>R(H \circ S)</math> is small then by the theorem, we can have good generalization performance. | |||
==Misc== | ==Misc== | ||