Dual-clustering-based Two-population Co-evolutionary Algorithm for segmentation coding in flash memory

Abstract

To ensure the reliability of stored data, error checking and correction (ECC) module is widely used in flash memory-based storage devices. With the increase of raw bit error rate (RBER), the traditional error correction coding mode is not only hard to satisfy the requirements of storage devices, but also reduces the efficiency of data recovery when error data cannot be located. In this paper, a novel segmentation coding mode is proposed to improve the drawbacks of traditional coding mode. To our knowledge, this is an innovative study of the error correction coding mode in ECC. It can make the coding mode used in ECC to become more flexible and efficient. Since the performance of this coding mode is related to segmentation, we propose a dual-population co-evolutionary algorithm based on clustering algorithm to optimize the performance. The proposed algorithm adopts clustering algorithm to measure the diversity of the population and dynamic weight allocation strategy to regulate evolution indicators of the population. Some experiments are conducted on benchmark problems and the segmentation coding problem, respectively. Experimental results show that the proposed algorithm is superior to other state-of-the-art evolutionary algorithms.