Scalable Distributed Genetic Algorithm for Data Ordering Problem with Inversion Using MapReduce

Logofatu, Doina and Stamate, Daniel. 2014. 'Scalable Distributed Genetic Algorithm for Data Ordering Problem with Inversion Using MapReduce'. In: AIAI 2014: 10th IFIP International Conference on Artificial Intelligence Applications and Innovations. Rhodes, Greece. [Conference or Workshop Item]

No full text available

Abstract or Description

We present in this work a scalable distributed genetic algorithm of Data Ordering Problem with Inversion using the MapReduce paradigm. This specific topic is appealing for reduction of the power dissipation in VLSI and in bioinformatics. The capacitance and the switching activity influence the power consumption on the software level. The ordering of the data sequences is an unconditional consequence of switching activity. An optimization problem related to this topic is the ordering of sequences such that the total number of transitions will be minimized – Data Ordering Problem (DOP). Adding the bus-invert paradigm, some sequences can be complemented. The resulting problem is the DOP with Inversion (DOPI). These ordering problems are NP-hard. We establish a scalable distributed genetic approach - MapReduce Parallel Genetic Algorithm (MRPGA) for DOPI, MRPGA_DOPI and draw comparisons with greedy algorithms. The proposed methods are estimated and experiments show the efficiency of MRPGA_DOPI.

Item Type:

Conference or Workshop Item (Paper)

Identification Number (DOI):

https://doi.org/10.1007/978-3-662-44654-6_32

Keywords:

Data Ordering with Inversion, Low Power, Distributed Algorithm, Evolutionary Approaches, Transition Minimization, Greedy, MapReduce, Apache Hadoop

Related URLs:

Departments, Centres and Research Units:

Computing

Dates:

DateEvent
14 June 2014Accepted
September 2014Published

Event Location:

Rhodes, Greece

Item ID:

11274

Date Deposited:

06 Feb 2015 14:44

Last Modified:

17 May 2019 10:01

URI:

https://research.gold.ac.uk/id/eprint/11274

Edit Record Edit Record (login required)