Improvements on the Johnson bound for Reed-Solomon codes

  • Authors:

  • V. N. Muralidhara;Sandeep Sen

  • Affiliations:

  • Department of Computer Science and Engineering, Indian Institute of Technology, New Delhi - 110016, India;Department of Computer Science and Engineering, Indian Institute of Technology, New Delhi - 110016, India

  • Venue:
  • Discrete Applied Mathematics
  • Year:
  • 2009

Quantified Score

Hi-index 0.04

Visualization

Abstract

For Reed-Solomon codes with block length n and dimension k, the Johnson theorem states that for a Hamming ball of radius smaller than n-nk, there can be at most O(n^2) codewords. It was not known whether for larger radius, the number of codewords is polynomial. The best known list decoding algorithm for Reed-Solomon codes due to Guruswami and Sudan [Venkatesan Guruswami, Madhu Sudan, Improved decoding of Reed-Solomon and algebraic-geometry codes, IEEE Transactions on Information Theory 45 (6) (1999) 1757-1767] is also known to work in polynomial time only within this radius. In this paper we prove that when k0) there can be at most O(n^c^2^@a^(^1^-^@a^)^^^2^+^c^+^2) number of codewords. For any constant c, we describe a polynomial time algorithm for enumerating all of them, thereby also improving on the Guruswami-Sudan algorithm. Although the improvement is modest, this provides evidence for the first time that the n-nk bound is not sacrosanct for such a high rate. We apply our method to obtain sharper bounds on a list recovery problem introduced by Guruswami and Rudra [Venkatesan Guruswami, Atri Rudra, Limits to list decoding Reed-Solomon codes, IEEE Transactions on Information Theory 52 (8) (2006) 3642-3649] where they establish super-polynomial lower bounds on the output size when the list size exceeds @?nk@?. We show that even for larger list sizes the problem can be solved in polynomial time for certain values of k.