Problems Solvable in Polynomial

Space
Alice Lewis
May 6, 2002

Polynomial Space Problems

Allow Turing machine M to use polynomial
amount of space in input, regardless of
amount of time needed
Consider deterministic as well as nondeterministic

Polynomial-space-bounded
Turing Machine
Finite
Control

--- input w-----

n cells
------cells ever used---------
p(n) cells

Polynomial Space Problems

polynomial p(n) given input w of length
n, the TM never visits more than p(n) cells
of its tape
By Theorem 8.12, studied previously, we
may assume that the tape is semi-infinite
and the TM never moves left of the first cell

PS and NPS
PS = class of polynomial space languages
Includes L(M) for some polynomial-spacebounded, deterministic TM M

NPS = class of non-deterministic

polynomial space languages
Includes L(M) for some non-deterministic
polynomial-space-bounded TM M.

Clearly, PS is contained in NPS

To Prove: PS=NPS
Theorem 11.3:
If M is a polynomial-space-bounded TM (DTM
or NTM), and p(n) is its polynomial space
bound, then there is a constant c such that if M
accepts its input w of length n, it does so within
c1+p(n) moves

To Prove: PS=NPS
Theorem 11.4
If L is a language in PS (or NPS), then L is
accepted by a polynomial-space-bounded
deterministic (nondeterministic) TM that halts
after making at most cq(n) moves, for some
polynomial q(n) and constant c > 1

To Prove: PS=NPS
The proof of P = NP is difficult to imagine;
PS = NPS is easy
Proof involves
Simulation of NTM that has a polynomial space
bound p(n) with a DTM with polynomial space
bound O(p2(n))
Uses recursive function reach described on
page 476 of text

To Prove: PS=NPS
Theorem 11.5: Savitchs Theorem
using previous theorems, we need only prove
that NPS PS
to prove this, we use the fact that the total
amount of space used for non-deterministic TM
is O(p2(n)), which is still polynomial

NP
PS=NPS

P
co-NP

Recursive