NeuroCOLT

Neural Networks and Computational Learning Theory

 

About NeuroCOLT

Papers Archive

1994 1995
1996 1997
1998 1999
2000 2001

Books

info@neurocolt.org

NeuroCOLT Technical Report NC-TR-94-008

 

Generalized Knapsack Problems

Felipe Cucker
Universitat Pompeu Fabra
Balmes 132, Barcelona 08008
SPAIN

Michael Shub
IBM T.J. Watson Research Center
Yorktown Heights
NY 10598
USA

Abstract

One weaker form of the P \not= NP problem consists of fixing the degree of the polynomial bounds for deterministic and nondeterministic time.  The problem obtained in that way, to determine whether DTIME(O(n^d)) = NTIME(O(n^d)) has been solved in the Boolean setting only for the special case d=1.  In the context of real Turing machines the question of whether nondeterministic time is more powerful than deterministic time for fixed degree polynomial time bounds also arises naturally. As a first result one can remark that the quoted result for the Boolean situation still holds in the real case since the Knapsack problem can be solved in linear nondeterministic time while it is known that it has an \Omega(n^2) lower bound for deterministic time. This paper shows that in this context a stronger result holds; namely, that for each d \geq 1 the inclusion DTIME(O(n^d))\subset NTIME(O(n^d)) is strict.

 

Download Compressed Postscript