ISIT2001, Washington, DC, June 24-29,2001 Matrix Characterization of Generalized Hamming Weights G. Viswanath B. Sundar Rajan’ Dept. of Elect. Comm. Engg. Indian Institute of Science Bangalore 560012, India e-mail: gviswaaprotocol . ece.iisc.ernet.in Dept. of Elect. Comm. Engg. Indian Institute of Science Bangalore 560012, India e-mail: bsrajanaece. i i s c . ernet.i n Abstract - A linear code with the systematic generator matrix [I I P ] is Maximum Distance Separable (MDS) if and only if every square submatrix of P is nonsingular. In this paper we obtain similar matrix characterization for all linear codes with a specified Hamming Weight Hierarchy (HWH). Using this we characterize Near-MDS codes (NMDS), NearNear-MDS ( N ’ M D S ) codes and a generalization of these codes called N p M D S codes in terms of their systematic generator matrices. 0 1. a n d i < g 5 min{di-l,k), every ( g - 2 i + l + p I g ) sub matrix has rank 2 (g - i 1). + 2. there exist a 9, i < g 5 m i n { d i , k } , such that (9 - 2i + p , 9) sub matrix has rank equal t o (g - i ) . 0 Corollary 2: (N’MDS code) For N 2 M D S codes the systematic generator matrix characterization is The support of a linear code C is the set of coordinate positions, where not all codewords of C are zero. The r-th generalized Hamming weight d,, 1 5 r 5 k , of a [ n , k ] linear code C over a field is defined as the cardinality of the minimal support of an [n,r] subcode of C. The sequence ( d l , &, . . . ,&) is called the Hamming Weight Hierarchy (HWH) of C [3]. The relation d, = n - k r for r = 1 , 2 , . . . ,k characterizes MDS codes. NMDS Codes: [l]The class of [n,k ] codes where d l ( C ) = ( n - k ) , and d i ( C ) is ( n - k i ) for i = 2 , 3 , . . . , k . N’MDS Codes: [2] The class of [n,k ] codes where d l ( C ) = ( n - k - l),& ( C ) = ( n - k 1) and di(C) is ( n - k i ) for i = 3 , 4 , . . . ]IC. We generalize NMDS and N’MDS codes as N p M D S Codes: The class of [ n , k ]codes where di(C) = (n- k 2i - p - 1) for i = 1 , 2 , . . . ,k . 1. For 1 < g 5 min{(n - k - 2), k } every (9 matrix has rank 2 g 3. For 1 < g 5 min{(n - k - l),k } there exits a (g sub matrix with rank (9 - 1). + + 5 . For 1 < g 5 min{(n - k ) , ( k - 2)) every (g,g matrix has rank 9. + + 0 0 + + 0 0 For i < g 5 min{di - 1,k } , every (9 +vi matrix of P has rank 2 ( g - i + 1). + 1 - i ,9) sub 0 For 1 < g 5 min{(n - k - l),k } every (9 matrix has rank 2 g + 1,g) sub For 1 < g 5 min{(n - k ) , k } there exists a (g,g) sub matrix with rank equal to (9 - 1) For 1 < g 5 min{(n- k ) , ( k - 1)) every (g,g matrix has rank 9. + 1) sub Corollary 3.1: If n > ( k + q ) the systematic generator matrix of a N M D S code is characterized by every (g+l, 9) submatrix having rank 2 g. There exists a g, i < g 5 min{di,k}, such that rank of (g - i vi,9) sub matrix is (9 - i) + REFERENCES F~~1 < 5 min{(n - k ) , ( k - cl)} every (9,9 + p ) sub matrix has rank 9. For N’MDS codes the defect1 ‘%I for 1 5 i 5 PI is ( p + - i). The defect is zero for all i > p . Corollary 1: ( N p M D S codes) The systematic G matrix characterization of N p M D S codes is as follows: ‘This work was partly supported by DST, India, through Research Grant No:III.5(31)/99-ET to B. S. Rajan 0-7803-7123-2/01/$I 0.00 02001 I EEE + 2) sub Corollary 2.1: For k > q > 3 and n > 2q - 1 k the systematic generator matrix of a N 2M D S code is characterized by every (g 2, g) submatrix having rank 1 g. Corollary 3: ( N M D S code) Systematic generator matrix characterization of N M D S codes is: 11. MATRIXCHARACTERIZATION OF HWH The M D S discrepancy of an [n,k] code is defined as the smallest p such that d,+l = n - k p 1. Theorem 1: The systematic parity check matrix of an [n k ] linear code with M D S discrepancy p and & ( C )= n-k+i-r/i can be characterized as follows: + 1,g) 4. For 2 < g 5 min{(n - k l),k } there exits a (g - 1,g) sub matrix with rank (g - 2). + + + 2,g) sub 2. For 2 < g 5 min{ ( n- k ) , k } every (g,g) sub matrix has rank 2 (9 - 1) + + F o r i = ( p + l ) , ... , k 1. For 1 < g 5 m i n { ( n - k ) , ( k - p ) } every ( g , g + p ) sub matrix has rank 9. I. PRELIMINARIES ‘. For i = 1,2, . . . ,p 61 [I] S. D. Dodunekov and I. N. Landgev, ’ On Near-MDS Codes’, Technical Report, No:Lath-ISY-R-1563, Department of Electrical Engineerinn, - Linkoping . -University, February, 1994. [2] J. Olsson, ’ On Near-Near-MDS Codes’, Proceedings of Algebraic and Combinatorial Coding Theory Workshop, June, 1996 [3] V. K. wei, ”Generalized Hamming Weights for linear codes,” IEEE Trans. on Information Theory, IT-Vo1.37, No:5, Sept.1991, pp.1412-1418. In