HEAL Down Under

February 3, 2012
Michael Affenzeller and quite a few HEAL members are currently on their way to Sydney, Australia, to attend the IEEE APCast'12 conference next week. APCast'12 will be hosted by the University of Technology (UTC), Sydney, Australia in conjunction with Universidad de Las Palmas de Gran Canaria, Spain. The conference will be an excellent opportunity to catch up with Zenon Chaczko from UTC, who visited us at the University of Applied Sciences Upper Austria a couple of years ago and in return invited Stefan Wagner as UTC guest lecturer in 2008. We will be back on 13. February.

HeuristicLab 3.3.6 Released

January 3, 2012
Happy new year! The HeuristicLab development team kicks off 2012 with the brand new HeuristicLab 3.3.6 release.

One of the most exciting new features of HeuristicLab 3.3.6 is HeuristicLab Hive which provides an infrastructure for parallel and distributed computation. Hive consists of a server with a database and a number of computation slaves. A user can upload a job to the Hive server which distributes the jobs among the available slaves and automatically retrieves the results after they are finished. While the main purpose of HeuristicLab Hive is to execute HeuristicLab algorithms, it can also be used independently. A detailed intro to Hive can be found in the Hive How-To on the HeuristicLab webpage.

Additional new features in HeuristicLab 3.3.6 include:
  • Robust Taboo Search for the Quadratic Assignment Problem
  • New Standard Algorithms for Regression and Classification (kNN, Neural Networks, Multi-Nominal Logit Regression)
  • Genetic Programming Grammar Editor
  • New Standard Tree-Creation Operators for Genetic Programming (Grow, Full, Ramped Half-Half)
  • Combination and Transformation of Algorithm Results
  • Customizable Charts
  • Performance Benchmarks etc.
For the full feature list, samples and documentation, go to http://dev.heuristiclab.com.

EvoSoft Workshop @ GECCO 2012

January 2, 2012
We are pleased to announce that Michael Affenzeller and Stefan Wagner will chair the Evolutionary Computation Software Systems (EvoSoft) Workshop at the GECCO 2012 conference. Stefan Wagner is the the project manager and head developer of the HeuristicLab optimization environment. The workshop will help to identify common efforts in the development of evolutionary computation software systems and will be a great platform to discuss cooperation potentials and synergies between different research groups.
GECCO 2012 will take place in Philadelphia, USA, from July 07-11, 2012.
Submission deadline: March 28, 2012 (hard deadline)
Notification of acceptance: April 9, 2012
Camera-ready submission: April 16, 2012
Workshops at GECCO 2012: July 7 - 8, 2012

Additional details are available at http://evosoft.heuristiclab.com and http://www.sigevo.org/gecco-2012.

Financial Analysis with HeuristicLab (Part 2)

December 23, 2011
Back in September Gabriel Kronberger posted about financial analysis over at the HeuristicLab Blog and showed some preliminary results for modelling the European interest rate swap yield with genetic programming. After many months of algorithm tweaking and refinement he presented the completed analysis and generated prognosis models to scientific peers at the ERCIM11 conference in London last weekend. ERCIM11 was organized by the London School of Economics, Queen Mary University of London and Birkbeck University of London.

Michael Affenzeller as Guest Lecturer in Las Palmas

December 20, 2011
Michael Affenzeller spent the better part of last week at the Universidad de Las Palmas de Gran Canaria, Spain. He was invited by Carmen Paz Suárez Araujo, the university’s vice head of internationalization, to hold a guest lecture in her PhD seminar and discuss future collaboration opportunities between our research groups.

Carmen Paz and her group conduct research in the area of artificial neural networks (ANN) with a special focus on (bio-)medical applications and hybrid algorithmic variants. We found that their research on modeling Alzheimer's Disease and related forms of dementia via ANN might tie in nicely with the heuristic dementia models developed by researchers around long-term HEAL member and bioinformatics research group head Stephan Winkler. Our groups are currently preparing a joint research proposal in that area, which will be submitted to the funding agency within the next couple of months. As usual, we will keep you posted about new developments on our homepage.

Erik Pitzer in Riga

December 14, 2011

Riga Technical University is a long-term scientific partner of our group. So far, two PhD students from Latvia have visited us in Hagenberg: Tatjana Lagzdina in 2008 and Vitaly Bolshakov in the summer of 2011.

During his stay in Austria, Vitaly and Erik Pitzer indentified a common research interests in the area of fitness landscape analysis, which led to a joint publication on vehicle scheduling problems. Due to the positive and fruitful collaboration, Erik is currently staying in Riga for a reciprocal visit. This time, they will tackle the vehicle routing problem and perform a rigorous fitness landscape analysis on representative problem instances. We will keep you updated about the results on our homepage, as usual.

2nd Workshop on Young Academics' Management Science

December 5, 2011

HEAL memeber Stefan Vonolfen went to Graz to attend the 2nd Workshop on Young Academics' Management Science last weekend. The workshop is organized each year by Dr. Peter Greistorfer and Dr. Marc Reimann from the University of Graz with the aim to encourage networking between young post-docs and grad students in the fields of metaheuristics and operations research. Stefan presented some recent research results in his paper "An Online Genetic Algorithm for Solving Dynamic Pickup and Delivery Problems".

Large Experiments in HeuristicLab

October 27, 2011
Andreas Beham recently had to run some large parameter experiments in HeuristicLab - and by large we mean about 400 configurations per experiment and more than 27,000 individual runs. The goal of this experiment was to test several parameter configurations on a number of problem instances in order to find the configuration that performed best in those instances.
Andreas shares his thoughts on the design of large parameter experiments and how you can make them as memory-friendly as possible on the HeuristicLab Blog. The blog entry also comes with a helpful code sample that should get you started, if you face a similar problem. Comments and questions are of course always welcome, either directly via mail or in our google group!

HeuristicLab Tutorial in Barcelona

October 27, 2011
During the IMMM 2011 a tutorial on "System Identification and Data Mining with HeuristicLab"
was held by Andreas Scheibenpflug and Michael Kommenda. In particular, they showed how to
  • generally work with HeuristicLab
  • setup and analyze large scale optimization experiments
  • perform regression and classification analysis and
  • avoid overfitting during the modeling phase.

The slides to this tutorial should be available on the conference homepage soon.


Two "newly minted" PhDs

October 7, 2011

We are very happy to announce that two long-term HEAL affiliates from the Institute for Production and Logistics Management at the Johannes Kepler University Linz have successfully defended their PhD theses. The orals were scheduled back to back on Tuesday with many friends and colleagues in the audience. Their individual research areas were:

  • Roland Braune: Job Shop Scheduling with Total Weighted Tardiness Objective: Enhancements to Local Search and Bottleneck-based Methods.
  • Michael Bögl: Suchraumstrukturanalysen und selbstlernende nachbarschaftsbasierte Suchverfahren am zeitfensterbeschränkten Problem des Handlungsreisenden

In both cases Prof. Zäpfel served as primary PhD supervisor and Michael Affenzeller agreed to be secondary supervisor. Roland and Michael will receive the title Doctor Technicae from the Johannes Kepler University Linz. The whole team congratulates to this success and wishes them all the best for their future careers!

Syndicate content