CosetAnalysisofReedMullerCodesViaTranslatesofFinite.pdfVIP

INFORMATIONAND CONTROL20, 410-414 (1972) Coset Analysis of Reed Muller Codes Via Translates of Finite Vector Spaces R. P. KURSHANAND N. J. A. SLOANE Bell TelephoneLaboratories, Incorporated Murray Hill, New Jersey 07974 In order to calculate the error probability of a code it is necessary to know the distribution of the coset leaders. The enumeration of cosets of the first order Reed Muller code has been studied (i) by Berlekamp and Welch (1972) by associating with each coset a class of Boolean functions, and (ii) by Sloane and Dick (1971) by associating with each coset a class of single-error-correcting codes called structure codes. It is shown here that it is possible to have two vectors of the same weight and in the same coset, which yet have inequivalent structure codes, at least for lengths / 128, thus giving a negative answer to a question raised by Sloane and Dick. INTRODUCTION The set of all binary vectors of length n forms an n-dimensional vector space Vn over GF(2), and we will often think of these vectors as coordinates of the vertices of a unit cube in n dimensions. Let V~* denote the set of nonzero vectors of Vs . A set of 2~ binary vectors of length n is called an (n, k) linear error-correcting code if it is closed under componentwise addition modulo 2. An important family of such codes are the (binary) simplex codes or shortened first-order Reed Muller codes (Wozencraft and Jacobs, 1965; Berlekamp, 1968a; Posner 1968). Any such code is a subgroup of F~ and so partitions V~ into cosets. The classification of these cosets