We are delighted to announce that our recent research work related to online model selection for constrained optimization has been accepted for publication and it will be presented at SSCI 2016. The SSCI 2016 conference will be held in Athens, Greece, from 6th to 9th of December, 2016. Many plenary and keynote talks are planned in the wide filed of computational intelligence.
Our article "Online Selection of Surrogate Models for Constrained Black-Box Optimization" received many positive reviews. In our former work, we have developed a surrogate assisted algorithm for optimization under constraints which builds a separate surrogate model for each constraint and objective function. Then it tries to solve the optimization problem on the surrogates. So far, we always used the same model type for all functions. Now, these functions can be of completely different types for each constraint and objective, leading to possibly better models. This formed our motivation to think about an online algorithm which selects out of an ensemble the best model for each function in each iteration. We have shown that the SACOBRA optimizer with ensembles of models improves success rates by 15% in comparison to SACOBRA with a fixed model.
The figure shows our best ensemble "MQ-Cubic" (an ensemble of different multiquadric RBFs and cubic RBFs, red curve), in comparison to fixed variants (blue and violet curves). The data profile is a measure of optimization success on a suite of problems, the higher the better:
If you are interested in reading our paper you can download it from http://www.gm.fh-koeln.de/~konen/Publikationen/Bagh16-SSCI.pdf. If you are attending the SSCI 2016 conference, don't miss our presentation .
This year the bi-annual IEEE World Congress on Computational Intelligence WCCI was held in beautiful Vancouver, Canada. WCCI is one of the biggest world congress on Computational intellignce and this year the number of participants with about 1800 researchers broke all the time record. You can find more information about this congress here.
Our team had the honor to present our recent work on "equality constraint handling" in this important conference. If you are interested in our work you can find it here. In case you are interested to have access to our presentation slides please contact us.
SAMCO workshop took place from February 29, 2016 till March 4, 2016 at the Lorentz Center in Leiden, The Netherlands. In this workshop many researchers and PhD students who are active in the field of surrogate assisted multi criteria optimization were invited. Among the participants there were three PhD students (Martin Zaefferer, Martina Friese, Samineh Bagheri) from Cologne University of Applied Sciences (TH Köln). But not all the participants were from academia and many were from industry.
The workshop's main concept was to share different ideas toegther and also work together on different subjects. During a week of workshop 5 subjects were discussed in 5 different sub-groups and some preliminary results were discussed, you can find more information about what was going on in this workshop here. Furtheremore, PhD students could find a chance to present some parts of their research work to the others and recieve some feedback. You can find many of the presentations here.
The 25th Workshop Computational Intelligence 2015, an annual conference held in Nov'2015 by the Computational Intelligence (CI) Chapter of VDI-GMA (Gesellschaft für Mess- und Automatisierungstechnik) in Dortmund, has attributed the Young Author Award to Samineh Bagheri, PhD, scientific member of the research group of Professor Konen at Campus Gummersbach. Congratulations!
The research group supervised by Prof. Wolfgang Konen from TH Köln has a collaboration with the Natural Computing group (head: Prof. Thomas H.W. Bäck) of the LIACS Computer Science department at Leiden University. On October,16th, 2015, a seminar took place in Leiden, Netherlands, in which both groups presented their recent research works.
This year GECCO took place in Madrid on 11-15 of July. Many interesting tutorials and workshops were presented in this conference. On 14th of July we presented our paper about a new repair method for constrained optimization. You can download the presentation slides from here.
We are glad to inform you that SACOBRA package is now available on CRAN.
SACOBRA package performs surrogate-assisted optimization by utilizing radial basis function interpolation. SACOBRA solver is a derivative-free optimizer appropriate to address constrained expensive optimization problems.
A new technical report on temporal difference (TD) learning for games and "self-play" algorithms for game-agent training is available. This report by Wolfgang Konen features a gentle introduction to TD learning for game play and gives hints for the practioner on the implementation of such algorithms . It shows the references to the most recent applications in this field and discusses in an appendix the more advanced topic of eligibility traces and how and why they work.
This report should be a help for people starting new in the field of TD learning for games and for people who work already in this field but struggle with specific details. It is an updated English translation of an earlier report in German language.
(2015): Reinforcement Learning for Board Games: The Temporal Difference Algorithm. Research Center CIOP (Computational Intelligence, Optimization and Data Mining) Cologne University of Applied Sciences, 2015.
Ms. Samineh Bagheri has won the 3rd price in the annual Erzquell award with her master thesis „Efficient Surrogate Assisted Optimization for Constrained Black-Box Problems“. My most cordial congratulations to her!
Many real-world optimization problems are dealing with constraints. The valid solution for constrained optimization problems (COP) lies somewhere in the feasible region which is a sub-set of the input-space. The borders of the feasible area are defined by one or many constraints. Existance of constraints makes the optimization problems more demanding. There are different approaches to handle the constraints but none of them can outperform all others for all different types of COPs.
There are only a few constraint handlers which work on the basis of fixing a good "infeasible solution" and generating feasible solutions by using the information coming from the infeasible candidates. We proposed a new technique to repair infeasible solutions. The correspoding work was accepted as a paper in GECCO 2015. A talk about this work will be given in GECCO conference on 11-15 of July in Madrid.