Log in

Log in


PhD-Reports sorted by


Year 2017

Year 2016 Year 2015 Year 2014

Year 2013 Year 2012 Year 2011

Year 2010 Year 2009 Year 2008



A1 A2 A3


B1 B3 B4

B5 B6 B7


C1 C3 C4

C6 C7

D2 D3 D5

D6 D7 D8


Z1 Z2 Z3




The Danish national museum, conference venue of EvoStar 2015.

Tanja Zerenner has received the IRTG Summer School Grant 2015 to attend EvoStar 2015, the leading European event on Bio-Inspired Computation (8-10 April 2015) in Copenhagen, Denmark.

EvoStar is a yearly conference covering diverse topics from the area of Evolutionary Compuation. More precisely EvoStar is comprised of four smaller 'sub-'conferences, namely EuroGP, the European conference on Genetic Programming, EvoCOP, the European conference on evolutionary computation in Combinatorial OPtimisation, EvoMUSART, the international conference on evolutionary and biologically inspired MUsic, Sound, ART and design and EvoApplications, covering applications of evolutionary computation to problems from biology, economics, image analysis, games and many more. This year's EvoStar was held in the city of Copenhagen, the Danish capital. The Danish national museum served as this year's conference venue.

EvoStar 2015 had approximately 170 participants mainly from European Countries; few also traveled from Canada, USA, New Zealand, Australia or South Africa.

For the main part the conference was split in three parallel sessions. Conference opening and closing were held as plenary sessions, each combined with an invited talk. The first invited talk on Wednesday was given by Pierre-Yves Oudeyer on 'Open-Source Baby Robots for Science, Education & Art'. The second invited talk was given by Paulien Hogeweg on 'Non-random random mutations: Evolution of Genotype-Phenotype mapping' focused on biological aspects. Together with her colleague Ben Hesper she had coined the term 'bioinformatics' in 1970 referring to the study of information processes in biotic systems. Bioinformatics is closely linked to the approach of Evolutionary Computation, as Evolutionary Computation is inspired by the information processes in biological systems. As it can already be guessed from the titles of the two invited talks, the topics covered by EvoStar are very diverse.


Largest room during panel discussion.  

Of special interest to me was the Panel Discussion on 'Future and Emerging Trends in Genetic Programming'. Panelists were among a few others, Wolfgang Banzhaf, author of the textbook 'An Introduction to Genetic Programming', William B. Langdon, coauthor of the open source book 'A Field Guide to Genetic Programming' and Sara Silva, author of the GPLAB package for Matlab, which I have built upon in my PhD work. To me the most interesting question discussed, was why in applied sciences Genetic Programming (GP) is in general much less popular than other learning methods, such as artificial neural nets (ANNs) for instance. The question was brought up by Sara Silva, who also provided various possible explanations. One important reason might be, that for many popular programming languages, GP is not contained in standard software packages for machine learning and statistics, while especially ANNs, but also algorithms for random forests or simulated annealing are often included. A second point might be, that to a given problem GP is likely to return a different solution each time it is applied, as GP is in great parts stochastic. As the GP output is readable code, this ambiguity can be immediately recognized by the users. However, getting a various number of solutions and not a single one can also be an advantage and can be used for ensemble approaches. It was pointed out, that the readability of GP output should be recognized as a big advantage compared to many other machine learning methods. For real world problems, models created by GP can be understandable and interpretable, which should be an important argument for at least testing GP based methods on real world problems, also in the geosciences, where to my experience up to know, ANNs are widely applied while GP is still little known.
For me, as a user of Genetic Programming, attending the EvoStar was definitely interesting and helpful. For people using GP or similar approaches, I can highly recommend attending on of the two large conferences on Evolutionary Computation (EvoStar or GECCO). For people interested, here just a few important things that differ from most geoscientific conferences:

  • Acceptance rates are often relatively low (about 35 %).
  • Each 'sub-'conference contains a 'best paper' session. The papers presented in these sessions are picked by the reviewers. The winner however is chosen by the audience by blind vote after the session. The 'best paper' sessions are typically the most crowded ones.
  • A conference contribution is peer-reviewed and in general equivalent to a journal publication. In informatics conference proceedings are a widely accepted option, if not even the common way, of publishing scientific articles. Work that has already been submitted to a journal must not be submitted to a conference.

Finally also a number of social events were offered by the organization committee including the conference reception, a conference dinner, a boat tour on the canals of Copenhagen and optional tours to the old town or Tivoli for instance.



The colorful houses of the Nyhavn (new harbor); view from a canal boat.



One of the many boats at Nyhavn; view from a canal boat.


PhD Student: Tanja Zerenner - Meteorological Institute, University of Bonn

PhD project (working title): "Atmospheric Downscaling using Genetic Programming"

Supervisor: Prof. Simmer

For further information, please contact:


tn328 54d609d0981ab

Nadine Horst
(geb. Heinrichs)

IRTG Coordinator

University of Cologne
Institute for
Geophysics and Meteorology

D-50923 Cologne
icon phone
icon fax
  +49 (0)221 470 1629
+49 (0)221 470 5161
icon mail   irtg@tr32.de

Precipitation radar Uni Bonn

Das Logo der Rheinischen Friedrich-Wilhelms-Universität; Copyright: Uni Bonn  

logo uni koeln 40x40

  logo rwth aachen 40x40   logo juelich   logo geoverbund small