WorksComputer Science
Hardness and Inapproximability of the Covering Radius for LDPC Codes
Abstract
Given a binary parity-check matrix and an integer , the covering-radius decision problem asks whether every attainable syndrome can be represented by an error vector of Hamming weight at most . We study this problem for sparse parity-check matrices with bounded row and column weights. Using a bounded-occurrence quantified-SAT reduction, we prove that the problem is -complete for matrices with maximum row and column weight three, and remains -hard under a constant multiplicative gap. Starting from Vertex Cover on 3-regular graphs, we also obtain an explicit NP-hard approximation factor under the same weight constraints. We remark that weight three is the exact threshold for hardness: if either the maximum row weight or the maximum column weight is at most two, the covering radius can be computed in polynomial time.