On sums of independent random variables with unbounded variance, and estimating the average degree in a graph

We prove the following inequality: for every positive integer n and every collection X1, . . . , Xn of nonnegative independent random variables that each has expectation 1, the probability that their sum remains below n + 1 is at least > 0. Our proof produces a value of = 1/13 0.077, but we conjecture that the inequality also holds with = 1/e 0.368. As an example for the use of the new inequality, we consider the problem of estimating the average degree of a graph by querying the degrees of some of its vertices. We show the following threshold behavior: approximation factors above 2 require far less queries than approximation factors below 2. The new inequality is used in order to get tight (up to multiplicative constant factors) relations between the number of queries and the quality of the approximation. We show how the degree approximation algorithm can be used in order to quickly find those edges in a network that belong to many shortest paths. 1 A new inequality For a rando...

• Type: Conference
• Year: 2004
• Where: STOC
• Authors: Uriel Feige
• Keywords: Algorithms, Independent Random Variables, Multiplicative Constant Factors, Random Variable X, STOC 2004