Extra-Mendelian Inheritnace and "Archival" Evolutionary Algorithms
Extra-Mendelian Inheritance We are exploring biologically inspired models of proposed in the plant Arabidopsis thaliana (Nature 434, 505-509), suggesting a "cache" of ancestral genomic information to help the repair process - acting as an extra-Mendelian inheritance mechanism. We have been using extra-Mendelian inheritance for 1) a simulated process of genetic repair 2) as an infrequent inheritance pathway to improve population diversity without negatively impacting population fitness.
Exploring an Evolutionary Hypothesis Our work makes two distinct contributions. First we are developing very efficient general-purpose algorithms for solving constraint problems (called GeneRepair and AncDE). Secondly, we are exploring the plausibility of the proposed repair mechanism through our computational experiments. These results show that including modest amounts of (archival or non-Mendelian) ancestral genetic information in an evolutionary process can reliably improve the solutions produced by evolutionary algorithms.
How did I get involved in this project?
In 2000 and 2003 I published two papers in collaboration with George Mitchell on a template driven mechanism for genetic repair. These papers showed that frequently changing the templates used for repair produced better results than using static templates.
In late 2005 or early 2006 George Mitchell gave a talk about this genetic repair work, when someone in the audience made him aware of the work on Arabidopsis thaliana by Lolly & Pruitt (2005).
In 2006 I extended the earlier work with George Mitchelel to include ancestral templates, under the expectation that these would produce worse results than our previous templates. This work was conducted as part of a final year undergraduate research-oriented project and produced somewhat confounding results. Sometimes the ancestral templates produce the worst results (as expected), but sometimes they produced the best results by a significant margin!
Results on that undergraduate work led to some research funding to explore these ancestral templates in greater detail, as part of a PhD project (Amy FitzGerald). Results showed that performing genetic repair using a cache of semi-recent ancestors (one cached ancestor per genetic/solution) produce better results than a wide range of comparably techniques. That work subsequently led to follow-on work with Donagh Hatton, Rushikesh Sawant and Siti Khadijah.
GeneRepair and AncDE Publications
- Varying Variants for AncDE with MDV between Target and Trial Vector Measurement , International Conference on Telecommunication, Electronic and Computer Engineering (ICTEC 2017),
Siti Khadijah Mohd Salleh, Diarmuid O’Donoghue, Abd Samad Shibghatullah, and Zuraida Bt Abal Abas, Melaka Malaysia, November 27-8, 2017 ().
- AncDE for Single-Objective Computationally Expensive Numerical Optimization With Most Appropriate Value Of Aup And Arp, Interdisciplinary ICT Practise Conference (IIPC 2017),
Siti Khadijah, Mohd Salleh, Diarmuid, O'Donoghue, Abdul Samad, Shibghatullah, Melaka Malaysia, November 6-8, 2017 ().
- Magnitude of Difference (MDV) between Target vector and Trial vector in AncDE and standard Differential Evolution, Proceedings of the 5th international conference on Artificial Intelligence, Computer Science & Information Technology (AICSIT 2017),
Siti Khadijah Mohd Salleh, O’Donoghue D.P, Abdul Samad Shibghatullah, pp 100-104, Pulau Pinang, Malaysia, 31st July & 1st August, 2017 (17 AICSIT).
- An Ancestor based Extension to Differential Evolution (AncDE) for Single-Objective Computationally
Expensive Numerical Optimization, IEEE Congress on Evolutionary Computation (IEEE CEC),
Sawant R, Hatton D, O’Donoghue D.P, pp 3228-3234, Sendia, Japan, May 2015 (15 AncDE, pre-print version). DOI: 10.1109/CEC.2015.7257293.
- Arabidopsis thaliana Inspired Genetic Restoration Strategies,
International Journal of Biometrics and Bioinformatics (IJBB), Volume 7: Issue 1, pp 35-48,
Donagh Hatton, Diarmuid P. O'Donoghue, 2013
(13-ijbb.pdf).
- Explorations on Template-Directed Genetic Repair Using Ancient Ancestors and Other Templates, The Genetic and Evolutionary Computation Conference (GECCO) - Constraint Handling Workshop,
Donagh Hatton, Diarmuid P. O'Donoghue, July 2011
(11-gecco.pdf).
- Genetic Repair Strategies inspired by Arabidopsis thaliana, Lecture Notes in
Artificial Intelligence (LNAI 6206), Amy FitzGerald, Diarmuid P. O'Donoghue and Xinyu Liu, pp 61-71, 2010
(10-lnai.pdf).
- Biologically Inspired Non-Mendelian Repair for Constraint
Handling in Evolutionary Algorithms, The Genetic and Evolutionary Computation Conference (GECCO)
Constraint Handling Workshop, Amy FitzGerald, Diarmuid P. O'Donoghue, Portland, Oregon, pp 1817-1825, 7-11 July 2010
(ACM Online,
10-gecco.pdf).
- Genetic Repair for Optimization under Constraints inspired by Arabidopsis thaliana
Parallel Problem Solving from Nature (PPSN),
Amy FitzGerald, D. P. O'Donoghue, Lecture Notes in Computer Science (LNCS 5199),
Springer Berlin / Heidelberg, Germany, pp 399-408, 2008
(Springer-online,
08-ppsn.pdf).
- GeneRepair - A Repair Operator for Genetic Algorithms The Genetic and Evolutionary Computation
Conference (GECCO), Chicago IL, July 2003, George G. Mitchell,
Diarmuid O'Donoghue, David Barnes, Mark McCarville (03-gecco.pdf).
- A New Operator for Efficient Evolutionary Solutions to the Travelling Salesman
Problem, Applied Informatics 2000, pp 771-774, 0-88986-280-X, 2000, George G. Mitchell,
O'Donoghue, D. Trenaman, A. (00-iasted-ga.pdf).
A jar.zip containing the class files for our AncDE used for CEC2015 are available (the .zip file should be 188,416 bytes in size on disk).
Past & present project members
George Mitchell, Amy Fitzgerald, Xinyu Liu, Donagh Hatton, Rushikesh Sawant, Siti Khadijah.
Back to Diarmuid's page.
URL: http://www.cs.nuim.ie/~dod/
Revised: 2013.
Contact: webmaster@cs.nuim.ie
Copyright 2000 National
University of Ireland, Maynooth.