A destructive evolutionary algorithm process

Authors:
Joe Sullivan;Conor Ryan
Affiliations:
University of Limerick, Biocomputing and Developmental Systems, Limerick, Ireland;University of Limerick, Biocomputing and Developmental Systems, Limerick, Ireland
Venue:
Soft Computing - A Fusion of Foundations, Methodologies and Applications - Special issue on Bio-inspired Learning and Intelligent Systems
Year:
2011

Citing 0
Cited 4

Optimising Flash non-volatile memory using machine learning: a project overview

Proceedings of the Fifth Balkan Conference in Informatics
How early and with how little data? using genetic programming to evolve endurance classifiers for MLC NAND flash memory

EuroGP'13 Proceedings of the 16th European conference on Genetic Programming
Estimating MLC NAND flash endurance: a genetic programming based symbolic regression application

Proceedings of the 15th annual conference on Genetic and evolutionary computation
Learning predictors for flash memory endurance: a comparative study of alternative classification methods

International Journal of Computational Intelligence Studies

Quantified Score

Hi-index	0.00

Visualization

Abstract

This paper describes the application of evolutionary search to the problem of flash memory wear-out. Flash memory differs from standard RAM in that it can wear out due to the manner in which it is programmed. The operating parameters, such as voltage levels, of flash memory are notoriously difficult to determine, as the optimal values vary from batch to batch. The current method in use is an expensive and time-consuming manual process of destructive testing. Understandably, this process is normally undertaken only at design time and testing on individual batches is normally not feasible. The results are sub-optimum solutions which do not minimise wear-out over the lifetime of the device. This is an enormously important issue in manufacturing, as most Flash Memory devices requiring reliability (e.g. solid state device disk drives) often have 100% or more redundancy to compensate for the wear-out rates. We establish the viability of a hardware platform that utilises an Evolutionary Algorithm to perform destructive experimentation on hard silicon in order to discover optimal or, at least favourable, operating parameter settings automatically in a manufacturing environment. Here, we describe this hardware and reveal results demonstrating an average life extension of between 250 and 350% over the factory set conditions with a maximum life extension exhibited of 700% all for cells within the same device over the factory settings. Furthermore, since the process is automated, it is possible to leverage the spread between process batches to further enhance device specifications, facilitating the near no-cost life extension of a split-gate flash memory device.