Pairwise Independence and Derandomization
Abstract: This set of notes gives several applications of the following paradigm. The paradigm consists of two complementary parts. The first part is to design a probabilistic algorithm described by a sequence of random variables so that the analysis is valid assuming limited independence between the random variables. The second part is the design of a small probability space for the random variables such that they are somewhat independent of each other. Thus, the analysis of the algorithm holds even when the random variables used by the algorithm are generated according to the small space.