Shortening of codes (Q798612): Difference between revisions

With the shortening of good large codes one obtains often best known shorter codes. One useful method was described by \textit{H. J. Helgert} and \textit{R. D. Stinaff} [IEEE Trans. Inf. Theory IT-19, 818--820 (1973; Zbl 0273.94014)]. This paper summarizes first some methods which were applied for shortening of codes, then describes a new one. Theorem 1: Let \(Z(n_ z,d_ z,N_ z)\) be a code (linear or nonlinear) and let \(l, w\) be integers with \(0<l<n_ z\), \(0\leq w<\min (1/2,d_ z/2)\). Then there exists for any \(n_ z-l\) positions a code \(R(n,d,N)\) with parameters \(n=n_ z-l\), \(d=d_ z-2w\), \[ N=\left\lceil N_ z\left(\sum^{w}_{i=0}\binom{l}{i} 2^{-l}\right)\right\rceil. \] A somewhat sharper analogous theorem is stated for linear codes, too. The parameters of, constructed this way, 167 codes are listed. Reviewer's remark: Theorems 1 and 2 are generally incorrect, it is possible only to prove \(d\geq d_ z-2w.\) Namely, let \(Z\) be the \((5,3,4)\)-linear code \(\{\underline 0;(1,0,1,0,1);(0,1,0,1,1);(1,1,1,1,0)\}\), \(l=2\) and \(w=1\). If one applies the above method for the 3rd and 4th positions then one obtains in any case a code with minimal distance \(d=2\).