An optimal machine learning classification model for flash memory bit error prediction

Barry Fitzgerald, Conor Ryan, Joe Sullivan

Research output: Chapter in Book/Report/Conference proceedingChapterpeer-review

Abstract

NAND flash memory is now almost ubiquitous in the world of data storage. However, NAND wears out as it is used, and manufacturers specify the number of times a device can be rewritten (known as program-erase cycles) very conservatively to account for quality variations within and across devices. This research uses machine learning to predict the true cycling level each part of a NAND device can tolerate, based on measurements taken from the device as it is used. Custom-designed hardware is used to gather millions of data samples and eight machine learning classification methods are compared. The classifier is then optimised using ensemble and knowledge-based techniques. Two new subsampling methods based on the error probability density function are also proposed.

Original languageEnglish
Title of host publicationStudies in Computational Intelligence
PublisherSpringer Verlag
Pages89-110
Number of pages22
DOIs
Publication statusPublished - 2019

Publication series

NameStudies in Computational Intelligence
Volume801
ISSN (Print)1860-949X

Keywords

  • Classification
  • Classifier ensemble
  • Error rate prediction
  • Flash memory
  • Machine learning
  • Subsampling

Fingerprint

Dive into the research topics of 'An optimal machine learning classification model for flash memory bit error prediction'. Together they form a unique fingerprint.

Cite this