IlliGAL Blogging

We've moved!

2006-10-17T16:18:00.000-05:00

We now host IlliGAL Blogging on our new website http://www-illigal.ge.uiuc.edu. Please go there for the latest blogging from the IlliGAL lab.

Thank you.

Press Release of Genalytics Model 6.0

2006-10-09T11:46:00.000-05:00

ThomasNet reports that Genalytics released Genalytics Model 6.0, that uses genetic algorithms to identify credit risks, anticipate customer attrition, and detect probable bankruptcies:

Genalytics Model 6.0 transforms large amounts of transaction data into predictive intelligence and uses genetic algorithms to automatically search and identify key combinations of time series variables found in any dataset. By studying time series patterns, organizations can identify credit risks, anticipate customer attrition, and detect probable bankruptcies. Software also allows users to format modeling output for Java and SPSS.

Doug Newell, founder of Genalytics, gave a talk at GECCO 2005, as reported here.

John Holland in Canada

2006-09-30T12:02:00.000-05:00

Mahantesh reports that John H. Holland is visiting St. John's, Newfoundland in Canada (here). Here is a synopsis of his talk on complex adaptive systems:

The Department of Computer Science hosts the world-famous computer scientist and inventor of Genetic Algorithms (GA's), Dr. John Holland, from the University of Michigan. His talk on Complex Adaptive Systems (CAS) will discuss the study of complexity, innovation, and relations between CAS and GA's. Sponsor: Computer Science.

No doubt this was the doing of Wolfgang Banzhaf, head of CS.

Embracing mistakes

2006-09-24T16:08:00.000-05:00

Sometimes it is important to make mistakes, lots of them, on the way to doing something new (here).

Genetic algorithm for Ruby

2006-09-23T05:50:00.000-05:00

Rubyforge has a fairly recent posting for a package called rGenetic:

rGenetic is a genetic algorithm package for Ruby. The idea is to develop an easily customizable suite, allowing simple integration into a problem requiring genetic algorithms or evolutionary programming.

Matthew Linnell is the project administrator. See here for additional detail.

Fast GA on Freshmeat

2006-09-23T05:43:00.000-05:00

Freshmeat has a new posting for a publicly available genetic algorithm called FGA (fast genetic algorithm). The code is by Alessandro Presta and claims to be

a simple yet powerful implementation of a general genetic algorithm, and provides many types of crossover and selection procedures. The library is easy to incorporate into other applications. Further improvements include parallelization of the algorithm in multi-processor environments and general performance optimizations.

See here for more detail.

Michael Corleone and the next generation GA?

2006-09-23T05:36:00.000-05:00

Read here about what The Godfather character and the next generation of genetic algorithms have to do with one another.

Some photos from Reykjavik

2006-09-21T11:15:00.000-05:00

Following Dave's suggestion, I'm posting some photos from Reykjavik, capital of Iceland, where PPSN 2006 was held. Iceland has about the same area as Portugal but with a population of only 300.000 people, which gives the lowest population density (2.7/km^2) for a european country.
About 2/3 of the population lives in or around Reykjavik.

Since I spent most of my time at Reykjavik, I only have photos from there, but Kumara and Pier-Luca have really nice photos of some great Icelandic landscapes. I hope they post some of them.

City Hall at Tjornin lake:

Tjornin lake:

The Sun Voyager by Jon Gunnar Arnason:

Laugavegur street (downtown main street):

Hallgrimskirkja church:

Reykjavik from Hallgrimskirkja church tower:

Blue Lagoon:

Bullet train designed by GA

2006-09-19T14:16:00.000-05:00

Kazu Inaba told me this morning that the new Shinkansen has been designed in part by a genetic algorithm. A quick google search confirmed the not-so-news here on the JR Central site:

"The development of the nose shape incorporated the latest analysis method (genetic algorithm*4), used in the development of airplane wings, for the first time in railway rolling stock. In doing so, we have developed a nose shape with the optimal aerodynamic performance accommodating 300 km/h operation, while ensuring the Series 700 seating capacity and spacious interior."

A picture of the new train is shown above.

GAs and fighting androids

2006-09-15T05:52:00.000-05:00

Vaughan at mindhacks.com has a nice post about Risto Miikulainen's NERO and NEAT here. NERO is a wargaming system in which the android soldiers adapt using neural net brains evolved using genetic algorithms.

Poster Sessions @ PPSN 2006

2006-09-14T19:31:00.000-05:00

PPSN had 7 poster sessions, each with 15 posters. Below are some pictures from the poster sessions.

Poster Session #7:

Shigeyoshi Tsutsui presenting cSA: Ant Colony Optimization with Cunning Ants:

Pier-Luca Lanzi presenting Automatic Test Pattern Generation with BOA:

Dirk Thierens presenting Exploration and Exploitation Bias of Crossover and Path Relinking for Permutation Problems:

Dependency-tree EDA in C++ online

2006-09-13T22:37:00.000-05:00

Just minutes ago I put online the C++ source code of the dependency-tree estimation of distribution algorithm (dtEDA), which is one of the simpler EDAs but it can encode dependencies between problem variables in the form of a dependency tree. The code is not restricted to binary strings (unusual for my code); each symbol in the string can have an arbitrary number of values (specified by a user-defined function). The code also includes bisection for determining the optimal population size for reliable convergence, which is a useful feature for obtaining scalability plots.

The main advantage of using dtEDA instead of more complex EDAs, such as BOA or ECGA, is that dtEDA uses simpler models and it should thus be faster. The disadvantage is that the models are restricted and the user won't get good scalability if the dependencies cannot be effectively encoded as dependency trees (and more complex structures are needed).

The main advantage of using dtEDA instead of simpler EDAs, such as PBIL or UMDA, is that dtEDA is capable of encoding dependencies between the variables and it does not assume that the variables contribute to the objective function independently. The disadvantage is that learning dependency trees is still slower than using the simple univariate distributions where the structure is fixed.

The code is heavily commented and documented (by hand and using Doxygen automatic documentation generation system). I hope you find it useful.

Downloads from MEDAL:

Biothing: GAs and generative architecture

2006-09-13T15:28:00.000-05:00

Omnithing has a nice post on biothing, a " transdisciplinary laboratory which research focuses on generative potential of physical and artificial computational systems for design.” Read more about biothing here.

PR at The Entrepreneurial Engineer

2006-09-13T15:21:00.000-05:00

Some public relations tips on press releases in the web world can be found over at TEE (here).

Flash: GECCO-2007 in London

2006-09-12T13:36:00.000-05:00

The Genetic and Evolutionary Computation Conference (GECCO)-2007) will be held in London, England, 14-18 July 2007 at University College of London. Keep an eye on the conference web site here for details as they become available.

GP/GA startup gets $2.5M round for electronics design tools

2006-09-12T08:11:00.000-05:00

Ann Steffora Mutschlerof EDN reports (here) that Solido Design Automation received a$2.5 million round of funding from noted genetic-programming expert John Koza, complexity guru Doyne Farmer, and others. The article does not disclose the nature of the tools used in Solido's systems, but company CTO Trent McConaghy has worked with genetic programming and multiobjective GAs in the past.

Linkage

2006-09-11T14:08:00.000-05:00

I've been remiss by not linking to sites that link to us. Codeprof (Try {Teach} Catch {Mistakes}) links to IlliGAL Blogging and his latest post wonders why info text books are so bad (here).

Tutorials Day at PPSN 2006

2006-09-10T18:33:00.000-05:00

Today’s schedule at PPSN included several tutorial talks. I had the opportunity to attend to Darrell Whitley’s tutorial, who gave a very good talk on practical guidelines for using evolutionary algorithms. Next, I went to the tutorial talk about the covariance matrix adaptation evolution strategy (CMA-ES), given by Nikolaus Hansen. I found the talk particularly interesting given the close connection with estimation of distribution algorithms.

A penguin goes to college

2006-09-10T17:14:00.000-05:00

Over the past months, I have departed from the usual genetic algorithms theme of this blog to report on my elder son's search for a college (see here, here, here, here, here, here, here, and here, and here). The long march of the penguin through childhood is over (see here to see a sample of the penguin's writing and here for a comic sample), and tomorrow he leaves the nest and goes off to the Medill School of Journalism at Northwestern University and the Public Affairs Residential College to seek his fame and fortune.

As an engineer, I'm supposed to be a dispassionate observer and designer, but as a father penguin, I must confess that I will miss this bird. Before leaving he penned a new cartoon for NU, something having to do with a baby wildcat and a penguin. Keep an eye on The Daily Northwestern for his byline.

Live from Reykjavik @ PPSN 2006

2006-09-09T18:19:00.000-05:00

Although I arrived late in the morning at Reykjavik I was able to attend the afternoon session of the Workshop on Empirical Methods for the Analysis of Algorithms, organized by Luis Paquete, Marco Chiarandini, and Dario Basso.

The talks raised interesting questions about the empirical comparison of metaheuristics, given the increasing number of new (and not so novel) algorithms proposed in the literature, day to day.
The workshop proceedings will be soon available here.

Tomorrow we will have several tutorial talks.

TEE press release gets picked up

2006-09-03T15:48:00.000-05:00

here.

The Entrepreneurial Engineer is available

2006-09-01T14:43:00.000-05:00

My new book is out and available. See here for more information.

Download Kandid

2006-08-28T13:04:00.000-05:00

Physics-envy passes along a link to Kandid, an interactive genetic algorithm that evolves art in response to user evaluations. Go here to download.

Hello to GE 531 students

2006-08-24T16:46:00.000-05:00

This entry is merely to say hello to those new GE 531 students (Genetic Algorithms) who are taking the course with me in class and online.

The first lecture is in the can, and the course is up an running. Still time to register for those interested.

EuroGP, Evo*, and Michael O'Neill

2006-08-23T13:13:00.000-05:00

Michael O'Neill writes to help plug next year's EuroGP and Evo* Conferences and Workshops:

We are drumming up some interest for EuroGP 2007 at the moment, and of course the two other associated conferences (EvoCOP and EvoBIO) and EvoWorkshops. It would be great if perhaps you might consider highlighting the events on your groups Blog?

I'm happy to put in the plug. The EuroGP wiki is here and the general Evo* website is here. The conferences are 11-13 April 2007 in Valencia, Spain. Tough duty.

Michael also notes his move to University College Dublin (tough duty) and the creation of the Natural Computing Research and Applications Group therein (see here). Please join me in wishing him the best of luck with this venture.