Honors Theory of Computation, Spring'99

Homework 5

-- DATE DUE: May 3, 1999
-- PROBLEMS:
  1. This simple exercise (suggested by Emily Chapman) tests your basic understanding of probabilistic computation. To determine if a binary number n is odd, the obvious algorithm is, of course, to determine the predicate LSB(n)=``the least significant bit of n is 1''. Consider the following (unusual) algorithm:
    LOOP:
    (1) Toss a coin. If head, then output LSB(n).
    (2) If tail, let n\ass n+2 and go back to LOOP.

    • (i) Let T be the complete computation tree on input n. Determine the least fixed-point valuation ValT.
    • (ii) Does this algorithm have bounded-error? zero-error?
    • (iii) Determine the running time t(m) of this algorithm on inputs of length m, assuming LSB(n) takes lg(n) steps to compute.
    • (iv) Determine the average running time t'(m) of this algorithm.
  2. Show that BPP is closed under union.
  3. Consider the following ``counting'' version of 3SAT:
    #SAT := { <F,k> : F is a 3CNF formula with exactly k satisfying assignments.} Prove that #SAT is complete for the class PP (under Karp reducibility).