This paper explores a mathematical optimization approach to error correction. Tanatmis et al. [2010] proposed an integer optimization formulation for the maximum likelihood decoding of binary linear codes. They developed a cutting plane algorithm in which valid inequalities for cutting off non-integral solutions are iteratively added to the continuous relaxation problem. By contrast, we focus on the integer optimization decoding; that is, we solve the integer optimization problem directly by using mathematical optimization software. We evaluate the computational performance of the integer optimization decoding and suggest its potential for practical use. The computational results demonstrate that in a lownoise condition, the integer optimization decoding achieved higher bit error rate performance than did the common sum-product decoding for low-density parity-check codes.
整数最適化復号法の性能評価
3092_0038_05.pdf
(1.63MB)
