R&D: Improvement of SP Decoding Considering Influence of Recording Patterns by Neural Network in SMR

IEEE Transactions on Magnetics has published an article written by Madoka Nishikawa, Yasuaki Nakamura, Graduate School of Science and Engineering, Ehime University, Matsuyama, Japan, Yasushi Kanai, Department of Engineering, Niigata Institute of Technology, Kashiwazaki, Japan, and Yoshihiro Okamoto, Graduate School of Science and Engineering, Ehime University, Matsuyama, Japan.

Abstract: “We study a low-density parity-check (LDPC) encoding and iterative decoding system for a shingled magnetic recording (SMR). In particular, we show the usefulness of applying a neural network in iterative decoding. We previously adopted the neural network to evaluate the log-likelihood ratio (LLR) related to row operations on the parity check matrix for the decoding target bit and improved the sum-product (SP) decoding. In this study, we propose to update the LLR considering the influence of noise depending on the recording pattern by providing the LLRs for the decoding target and its adjacent bits to the neural network in SP decoding. In addition, we explore the parameters to update the LLRs by applying the genetic algorithm (GA) to achieve more effective iterative decoding.“