5,321
edits
Machine Learning: Difference between revisions
→Theory
(→Theory) |
|||
| Line 168: | Line 168: | ||
====Theory==== | ====Theory==== | ||
[https://nowak.ece.wisc.edu/SLT09/lecture8.pdf Reference]<br> | |||
In the case where the Hypothesis class <math>\mathcal{H}</math> is finite, we have | |||
* <math>|L_D(h) - L_S(h)| < \sqrt{ | |||
\frac{\log|\mathcal{H}| + \log(1/\delta)}{2m}} | |||
</math> | |||
: where <math>m</math> is the size of the sample. | |||
For all h in H, | For all h in H, | ||
* <math>|L_D(h) - L_S(h)| < K_1 \sqrt{ | * <math>|L_D(h) - L_S(h)| < K_1 \sqrt{ | ||
\frac{VCdim + K_2 log(2/\delta)}{2m}} | \frac{VCdim + K_2 \log(2/\delta)}{2m}} | ||
</math> | </math> | ||
: for some constants <math>K_1, K_2</math> | |||
===Bias-Variance Tradeoff=== | ===Bias-Variance Tradeoff=== | ||