Another manifestation of the curse is that the sampling density is proportional to $N^{1/p}$, where $p$ is the dimension of the input space and $N$ is the sample size. Thus, if $n = 100$ represents a dense sample for a single input problem, then $n=100^{10}$ is the sample size required for the same sampling density with $10$ inputs.`

This is from Elements of Statistical Learning... Shouldn't $n=N^p$, isn't formula wrong?

  • 2
    $\begingroup$ No, the formula is correct. For a fixed number of points, the sampling density decreases as the dimension increases. $\endgroup$ – Matthew Drury Aug 5 '17 at 15:19
  • $\begingroup$ Oh, I was thinking about it in wrong way. So because distribution density decreases we need need more(100^10)inputs for it to be proportional to one dimensional estimation? $\endgroup$ – econ Aug 5 '17 at 15:25

Don't know anything about this topic but, if the sample density is 100 and it is proportional to $N^{1/p}$, then you have: $$ 100 = \alpha N^{1/p} $$ which implies $$ \left(\dfrac{100}{\alpha}\right)^p = N $$ So if the constant of proportionality is 1, the required sample size is $100^{10}$.

| cite | improve this answer | |

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.