Kenneth Alexander proved a uniform Law Of Iterated logarithm for Vapnik-Chervonenkis classes in the article Probability Inequalities for Empirical Processes and a Law of the Iterated Logarithm (Ann. Probab. volume 12, number 4 (1984), 1041-1067).

It states that given a VC class of sets $\cal C$ and a sequence of i.i.d. random variables $(Y_n)_n$, the following holds: $$ \limsup_{n \rightarrow \infty} \sup_{C \in \cal C} \frac{\left|\sum_{i=1}^{n} 1_{C}(Y_i) - nP_{Y_1}(C)\right|}{\sqrt{ 2 n \log \log n}} = \sup_{c \in \cal C} \left(P_{Y_1}(C)(1-P_{Y_1}(C))\right)^{1/2} $$

What if one has a slightly more general setting: the variables $(Y_n)_n$ 'live' on a general measure space $(\Omega, {\cal F}, \mu)$, where $\mu$ is a finite measure. Which conditions on the measure $\mu$ and the variables $(Y_n)_n$ should be imposed for the statement above to hold?

Best regards


Your Answer

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

Browse other questions tagged or ask your own question.