This preview has intentionally blurred sections. Sign up to view the full version.
View Full DocumentThis preview has intentionally blurred sections. Sign up to view the full version.
View Full DocumentThis preview has intentionally blurred sections. Sign up to view the full version.
View Full DocumentThis preview has intentionally blurred sections. Sign up to view the full version.
View Full Document
Unformatted text preview: Pairwiee Independence and Derandemizatien Michael Lubyr 3.: Avi Wigdersen Feundetiuns and Trends in Thenre‘tical Computer Science EDGE 0 Introduction to Pairwise Independence o Generating Pairwise Independent Bits
0 Introduction to l Derandomization o Brute Force Derandomization
o Derandomization via Pairwise Independence Pairwise Independent Events 0 Consider the set of events A1, An 0 The events A1, ...,A,, are mutually independent if
Pr[A1 0 A2... 0 An] : Pr[A1]Pr[A2]...Pr[A,,] o The events A; and A1 are pairwise independent if PrlAi ﬂ Aj] : Pr[A,]Pr[AJ] for all I,j Example of Pairwise Independent Events 0 Consider the set 0 2 {abc, acb, aaa, bac, bca, bbb, cab, cba, ccc}
and probability of any of them being picked is 5 e Let event Ak denote that the kth letter is a. Clearly Pr[Ak] : § 0 Pr[A1 0 A2] = Pr[A2 0 A3] = MM; 0 A1] = % o Pr[A1 0 A2 0 A3] = :15 75 Pr[A1]Pr[A2]Pr[A3] 0 Thus, the events pairwise independent but not mutually independent Pairwise Independence is weaker than Mutual Independence Generating Pairwise Independent Bits 0 We know how to generate mutually independent bits (Coin Flips) 0 Can we use mutually independent bits to generate pairwise
independent bits? Construction of n pairwise independent random bits with uniform distribution from a small sample space 0 For given n, let t = Iog(n+ 1) and let I = {1,..., t}.
0 Let random bits b1, ..., b, be obtained by tosses of a fair coin.
0 Let J1, ..., J,, be pairwise distinct nonempty subsets of I. o For‘i = 1, ..., n let r,~ : €919.13} . Observe that there are 2‘  1 pairwise distinct nonempty subsets of I Generating Pairwise Independent Bits Lemma: Verification of the construction o The random bits r1, ..., I), obtained by the construction above are
pairwise independent and each bit is uniformly distributed. 0 Fix any nonempty set J g {1, ..., t} and let r = @jejbj
o For any ﬁxed index 5 6 J we have, r = ejerj = ($jeA{s}l),)\bs o If the bj's have already been determined for all j e J. then
depending on the value of b5, the value of r will either agree with or
will differ from @jéerj 0 But b, attains its two possible values with probability; — each and the
b are mutually independent hence r will be uniformly2 distributed In {0 1} Generating Pairwise Independent Bits Proof(continued) 0 Consider any two set S, T(S ;£ T) g [t]. We have: R5 = Rsm‘ 69 R$\T
RT = RSnT 6 RT\$ o R5\T, RSnT and RT\5 are independent as they depend on disjoint
subsets of the bi's 0 At least two of these subsets are nonempty (because 5 gé T). This
implies that (R5, R7) each take value in {0, 1} with probability %. o The random bits generated are pairwise independent bits Brute Force Derandomization o Brute Force Derandomization refers to simulating a randomized
algorithm for every possible value of its random source. 0 Consider a randomized algorithm that runs for poly(n) steps on all
inputs of size n. o The algorithm might uses r random bits o Once the random bits are known, each execution can be seen as a
deterministic algorithm 0 There are 2’ possible combination of random bits 0 Algorithm: Execute randomized algorithm for each possible
combination 0 Time Complexity of deterministic algorithm 2'poly(n) steps
0 Done? No! Brute Force Derandomization c We want efﬁcient deterministic algorithm
0 2’poly(n) must be 0(n‘) for some constant c
o :> r : 0(Iogn) 0 Thus, when r = 0(Iogn), then even brute force derandomization
works well 0 First transform a randomized algorithm into one that uses only few
random bits, then apply brute force derandomization ...
View
Full Document
 Fall '10
 Staff
 Algorithms

Click to edit the document details