Error-correction of linear codes via colon ideals (Q497727)

Given an \((n,k,d)\)-linear code \(C\) over a field \(\mathbb{K}\) and a received word \(w\in \mathbb{K}^n\) with \(t\) positions are in error. The problem of syndrome decoding i.e., finding a codeword \(v \in C\) that is the nearest neighbor to \(w\) (i.e., \(v\in C\) that minimizes \(wt(w-v)\)) is of great interest in coding theory. In this paper, the authors take a more theoretical approach to understand what it means to error-correcting a received word through any linear code from an abstract point of view. The strategy to correct the errors in a received codeword is divided into two steps, first the received word \(w\) is added as a new row in the generating matrix of the code \(C\) which will result in an augment code \(C^{w}\). In the case where \(t\leq d-1\), the errors are the codewords of minimum weight \(d_w\) of \(C^w\). Second step, to determine these codewords (errors), the authors use tools and technique from commutative algebra such as primary decomposition of saturation of certain homogeneous ideal (generated by product of linear forms from the generating matrix of the code). Each prime ideal in the primary decomposition define a point in \(\mathbb{P}^{k-1}\) with the homogeneous coordinates give the coefficients in the linear combination of the row of the generating matrix that equals the codeword of minimum weight. In the good case, where we have only one nearest neighbor of \(w\). the saturation of the homogeneous ideal can be computed directly by coloning the ideal by a power of a single variable. Moreover, the authors determine the smallest such power for any linear code to be \(d_w\).