We investigate a novel approach for the iterative decoding of Trellis-Constrained Codes (a super-class of error-correction codes that comprises Turbo-codes and LDPC codes) which is inspired by the amplitude amplification principle used in quantum computing. The idea is to increase the probability of the correct word in every iteration, so that it will eventually stand out from all other words. We propose an approach that is provably converging, but the amplification factor can become so small that the decoding process practically stalls. We show that by heuristically retriggering the decoding process one can achieve an error-correcting performance that is similar to that of loopy belief propagation.
from cs updates on arXiv.org http://bit.ly/2UZ90tQ
//
0 comments:
Post a Comment