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-98-004


Are Lower Bounds Easier over the Reals?


Herve Fournier and Pascal Koiran
ENS Lyon

Keywords:Algebraic complexity, real Turing machine

Received: 04-MAR-1998


Abstract
We show that proving lower bounds in algebraic models of computation may not be easier than in the standard Turing machine model. For instance, a superpolynomial lower bound on the size of an algebraic circuit solving the real knapsack problem (or on the running time of a real Turing machine) would imply a separation of P from PSPACE. A more general result relates parallel complexity classes in boolean and real models of computation. We also propose a few problems in algebraic complexity and topological complexity.

Download Compressed Postscript