Evostar 2016

The Leading European Event on Bio-Inspired Computation.

Porto, Portugal, 30 March - 1 April 2016


Abstract: This talk falls into three main parts (as recommended by Aristotle). First, a look back at the heady days of the 1980s when evolutionary computing had an appealing air of raffish unorthodoxy at a time of peak Prolog. This will include a biased perspective on the genealogy of Genetic Programming, summarized in a simple spreadsheet as a staircase of synthetic serendipity, which I intend to use to explore the issue of what exactly are the defining features of GP, and who decides. Then I plan to discuss some of the design choices that arose in reviving and revising BEAGLE, in the light of experience with earlier versions. Some of these choices were guided by experimental trials, whose results will be reported. This reporting will include coverage of a couple of apparently neat ideas that don't seem to work. Finally, I will look ahead to the evolution (sorry, hard to resist sometimes) of the field. This provides an opportunity to indulge in some crystal-ball-gazing at indistinct impressions of the cybernetworked dystopia awaiting our grandchildren if the big-data fanatics ever get their acts together -- while still leaving time for questions from the audience.

Bio: Richard Forsyth is a relic from the medieval period of evolutionary computing. In 1981, inspired by even earlier pioneers like Oliver Selfridge and Gordon Pask, he published an account of what was arguably the first working example of tree-structured program code optimized by evolutionary methods. Nowadays it would probably be called Genetic Programming. At the time he called it a Darwinian rule-learner trained by "naturalistic selection". The term didn't catch on. In 1985 he started selling a PC version of the BEAGLE rule- finder system (Biological Evolutionary Algorithm Generating Logical Expressions). Fortunately his lack of business acumen spared him from the fate of becoming a software billionaire. In consequence, he has spent most of the last 30 years in the lower echelons of Britain's "higher" education system. His peregrinations have taken him on a tour through a motley assortment of faculties, including Arts, Medicine, Science and Social Sciences. Released from the academic hamster-wheel, he has recently revived & revised BEAGLE, with a few novel twists that will be described in his talk. This activity has caused BEAGLE to give birth to a rather mischievous pup called RUNSTER (Regression Using Naturalistic Selection To Evolve Rules). Both these systems will soon be made freely available to all & sundry on Richard's website.

Metaphors in metaheuristics - a symptom of a deeper ailment?

Abstract: Consensus is forming in (a large part of) the metaheuristics community that there is a problem with the current flood of "novel" metaphor-based methods. In the paper "Metaheuristics - the metaphor exposed", I have investigated why the field of heuristic optimization is especially vulnerable to this kind of bad science. In this talk, I further investigate this issue and arrive at the conclusion that the entire field of heuristic optimization is in need of an update of its standards.

Bio: Kenneth Sörensen (Antwerp, Belgium, 1974) holds a PhD from the University of Antwerp, obtained in 2003. Currently, he is a Research Professor of the Faculty of Applied Economics of the University of Antwerp. Within this Faculty, he founded the ANT/OR research group, that focuses on applications of operations research. Kenneth Sörensen has published a large number of articles in international refereed journals, and has presented his work at numerous scientic conferences. His main research interests are the application of advanced (metaheuristic) optimization methods and the development and study of optimization methods. Kenneth Sörensen is the founder and current coordinator of the EURO working group EU/ME – the metaheuristics community, the largest online platform for researchers in metaheuristics worldwide. He is also associate editor of the Journal of Heuristics, International Transactions in Operational Research, and 4OR.

Important dates:

Submission Deadline: 1 November 2015
EXTENDED DEADLINE: 11 November 2015
(site remains open for final changes until 15 Nov)
Notification: 4 January 2016
Camera-ready: 18 January 2016
Mandatory registration per paper: 15 February
Student bursary deadline: 20 February
Early registration discount: 25 February 2016
Registration deadline: 24 March
EvoStar dates: 30 March - 1 April 2016