Computational Complexity Endterm Exam Questions, Study notes of Computer Science

Six questions from a computational complexity exam. The questions cover topics such as decision problems for dnf formulas, the acceptance of certain families of circuits, multilinear polynomials, clique size in graphs, and quantum query models. The solutions are due by 5:00 pm on april 30.

Typology: Study notes

Pre 2010

Uploaded on 08/05/2009

koofers-user-79g
koofers-user-79g 🇺🇸

10 documents

1 / 3

Toggle sidebar

This page cannot be seen from the preview

Don't miss anything!

bg1
Endterm
CS6520
Computational Complexity
Solve all 6 questions. The solutions are due by 5:00 pm on Monday, April
30. Some hints are given on page 3. All the best!
Problems
1. A DNF formula in (boolean) variables x1, x2, . . . , xnis of the form
φ=D1D2· · · Dm
where for 1 im, Di=yi1yi2. . . yik
and each yjis a variable or its negation. Show that deciding if a DNF formula is
satisfiable is in Pbut counting the number of satisfying solutions is #P-complete.
2. Let Lbe the language accepted by a family of circuits {Cn}which consist of AND,
NOT and PARITY gates such that
Circuit Cnhas ninputs, size 2nO(1) and depth O(1).
AND gates have fan-in bounded by poly(n).
PARITY gates have unbounded fanin.
The circuits Cnare uniformly generated by a polynomial time DTM M.
Show that L P. In other words show that there is a polynomial time NTM N
which has an odd number of accepting computations on input xiff xL.
3. Let Z3={0,1,1}be the field of integers modulo 3. We say that a polynomial
P(X1,· · · , Xn) in nvariables is multilinear if the degree of each Xiin Pis at most
1. For instance P(X1, X2, X3) = X1X2+X2X3is multilinear but X2
1+X2
2is not.
Show that every function f:{0,1}nZ3is computed by a unique multilinear
polynomial in Z3[X1,· · · , Xn].
Consider all Boolean functions f:{0,1}n {0,1}. Let the degree of function f
be the degree of the unique polynomial computing f. Show that AND and OR
functions have degree n.
1
pf3

Partial preview of the text

Download Computational Complexity Endterm Exam Questions and more Study notes Computer Science in PDF only on Docsity!

Endterm

CS

Computational Complexity

Solve all 6 questions. The solutions are due by 5:00 pm on Monday, April

  1. Some hints are given on page 3. All the best!

Problems

  1. A DNF formula in (boolean) variables x 1 , x 2 ,... , xn is of the form

φ = D 1 ∨ D 2 · · · ∨ Dm

where for 1 ≤ i ≤ m, Di = yi 1 ∧ yi 2... ∧ yik and each yj is a variable or its negation. Show that deciding if a DNF formula is satisfiable is in P but counting the number of satisfying solutions is #P-complete.

  1. Let L be the language accepted by a family of circuits {Cn} which consist of AND, NOT and PARITY gates such that - Circuit Cn has n inputs, size 2n O(1) and depth O(1). - AND gates have fan-in bounded by poly(n). - PARITY gates have unbounded fanin. - The circuits Cn are uniformly generated by a polynomial time DTM M.

Show that L ∈ ⊕P. In other words show that there is a polynomial time NTM N which has an odd number of accepting computations on input x iff x ∈ L.

  1. Let Z 3 = { 0 , 1 , − 1 } be the field of integers modulo 3. We say that a polynomial P (X 1 , · · · , Xn) in n variables is multilinear if the degree of each Xi in P is at most
    1. For instance P (X 1 , X 2 , X 3 ) = X 1 X 2 + X 2 X 3 is multilinear but X 12 + X 22 is not.
      • Show that every function f : { 0 , 1 }n^ → Z 3 is computed by a unique multilinear polynomial in Z 3 [X 1 , · · · , Xn].
      • Consider all Boolean functions f : { 0 , 1 }n^ → { 0 , 1 }. Let the degree of function f be the degree of the unique polynomial computing f. Show that AND and OR functions have degree n.
  • The MOD-k function is 1 if

∑n i=1 xi^ is divisible by^ k, and 0 otherwise.^ Show that MOD-2 (PARITY) has degree n but MOD-3 has degree 2.

  1. Let ω(G) denote the size of the largest clique in graph G. Assume that there is a polynomial time reduction A that takes as input a SAT instance φ and outputs a graph G on n vertices such that - If φ is satisfiable, ω(G) ≥ αn. - If φ is unsatisfiable, ω(G) ≤ βn.

Here α, β are constants such that 0 < β < α < 1. Use this to show that, for any constant C, there is no polynomial time algorithm that approximates ω(G) within a factor C unless P = NP.

  1. Assume that there is an unknown Boolean function f : { 0 , 1 }n^ → { 0 , 1 } which is 1 at exactly K inputs. Give an algorithm to find (some) input x with f (x) = 1 which asks O(

N/K) queries in the Quantum Query Model (N = 2n). A single query Q is defined as the unitary operator:

Q |x〉 = (−1)f^ (x)^ |x〉 ∀x ∈ { 0 , 1 }n.

  1. The set-disjointess function f : { 0 , 1 }n^ × { 0 , 1 }n^ → { 0 , 1 } is defined as

f (x, y) = 1 ⇐⇒ xi ∧ yi = 0 ∀ i = 1, · · · , n

In other words, think of x and y as incidence vectors of sets S(x) and S(y) respectively. Then f (x, y) = 1 iff the sets S(x) and S(y) are disjoint. Let Mf denote the matrix of values of f.

  • Show that any 1-monochromatic rectangle in Mf has size at most 2n.
  • Show that the deterministic communication complexity of f is Ω(n).