On the Complexity of Estimating the Effective Support Size

Webpage for a paper by Oded Goldreich


Abstract

Loosely speaking, the effective support size of a distribution is the size of the support of a distribution that is close to it (in totaly variation distance). We study the complexity of estimating the effective support size of an unknown distribution when given samples of the distributions as well as an evaluation oracle (which returns the probability that the queried element appears in the distribution). In this context, we present several algorithms that exhibit a trade-off between the quality of the approximation and the complexity of obtaining it, and leave open the question of their optimality.

Material available on-line


Back to either Oded Goldreich's homepage or general list of papers.