Monday, July 31, 2006

 

Calumny halted at IB

I was recently forced to start moderating comments on this blog. Attacks of blog spam were stopped largely by requiring word verification (captchas), but offensive, ad hominem attacks have persisted to the point where all comments by non-members will be moderated before they are posted.

In general, comments will be posted (1) if they are relevant to the post they comment upon and (2) if they avoid slurring the author of the post and others in the community. IlliGAL Blogging has a point of view, and we welcome comments from those who disagree with us; controversial comments are the spice of blogging life, and the dialectic of online discourse can be enlightening. Having said this, we see no need to host comments that gratuitiously cast aspersions.

Of course, those who wish to continue their name-calling ways may choose to do so in the relative obscurity of their own blogs, but the pages of this blog are now closed to that type of comment.

Thursday, July 27, 2006

 

EDAs and cross-entropy methods

Estimation of distribution algorithms (EDAs)---also known as probabilistic model-building genetic algorithms (PMBGAs) and iterated density estimation algorithms (IDEAs)---are population-based optimization methods that use probabilistic modeling to guide the search for the optimum. In other words, we may say that EDAs replace standard variation operators of genetic and evolutionary algorithms by building and sampling a probabilistic model of promising solutions found so far. EDAs date back to 1994 when Shumeet Baluja proposed PBIL algorithm, and as most of us already know, since then the field has made a significant progress.

A similar approach has been taken in 1997 by Reuven Rubinstein, who proposed the cross-entropy method to provide an efficient tool for rare-event simulation and optimization. Just like EDAs, CE proceeds by generating samples of points from a probabilistic model and then updating the probabilistic model to better fit high-quality points in the generated sample. For me the most important question is: what can we learn from each other?

 

Even more genetic art

Jon McCormack has samples of his work on GEC-based art on his web page. Those who have looked at the first issues of SIGEVOlution (ACM SIGEVO newsletter) may notice that Jon was the one who supplied artwork for the cover.

Tuesday, July 25, 2006

 

More genetic art

IlliGAL Blogging reader and blogger Harold Skolnick (AKA www.physics-envy.blogspot.com) writes to show us some of his genetic algorithm art (see here). Also check out his photos (also here and here), drawings, paintings, and his blog.

 

MEDAL gathering: Post-event thoughts

MEDAL gathering on evolutionary computation is now over. I think that the speaker lineup was great, and the speakers have done an excellent job at presenting their recent results and new ideas; we should soon have PDFs of most presentations online at the event's web site. I was happy to meet a number of friends, some of whom had to drive for hours or even take several flights to get here. Overall, I think that the event was a success and I thank all our speakers and other participants for that. Soon we will put online a gallery of photos from the event, the list of participants, and, as I said earlier, PDFs of the presentations.

Monday, July 24, 2006

 

Mark Jakiela on open collaborative design of tangible artifacts

The last talk of MEDAL gathering is by Mark Jakiela and he is talking about open collaborative design of tangible artifacts and the role that evolutionary algorithms can play.

 

Shigeyoshi Tsutsui on EDAs for permutation problems

Shigeyoshi Tsustui is giving a talk on comaprison between probabilistic model building genetic algorithms with edge histogram vs. node histogram in permutation domain.

 

Moshe Looks on substructural learning in genetic programming

Moshe Looks talked about estimation of distribution algorithms for learning substructures in genetic programming.

 

Medal Gathering: Representations and heuristics in GP

Cezary Janikow is now presenting about representations and heuristics in GP.

 

Little models and big results

Dave Goldberg gave a very nice talk on the power of "little" models. Below is a slide in which he recalls about first meeting Martin.

 

MEDAL gathering: Kumara Sastry on efficiency enhancement of EDAs

Kumara Sastry is talking on efficiency enhancement techniques in estimation of distribution algorithms (EDAs).

 

MEDAL gathering: Kenneth Turvey on XCS in dynamic enviornments

Kenneth Turvey, a MEDAL member is talking about XCS in dynamic environments.

 

MEDAL gathering: Current and planned projects at MEDAL

Martin Pelikan is talking on the current and planned projects at the Missouri estimation of distribution laboratory (MEDAL).

Sunday, July 23, 2006

 

Getting ready for MEDAL gathering

I just arrived at Martin's place after 3 hours drive from Champaign. Martin Pelikan and Shigeyoshi Tsutsui are getting their presentations ready for MEDAL gathering tomorrow.


Thursday, July 20, 2006

 

Recent visits by country

Here are the latest stats of IlliGAL Blogging visitors by country:

  1. United States, 140
  2. Germany, 29
  3. Canada, 15
  4. United Kingdom, 13
  5. Italy, 10
  6. Netherlands, 7
  7. Austria, 7
  8. Japan, 7
  9. Portugal, 7
  10. Poland, 6

Canada beating our the UK is a bit of a surprise, and Poland would be a surprise, but Polish is one of two languages that Genetic Algorithms in Search, Optimization, and Machine Learning has been translated into (see here). The other is French.


 

Linkorama

It has been awhile since I've acknowledged those who link to us. Thanks to www.robsyvertsen.org, www.togelius.blogspot.com/, http://philosophyofengineering.blogspot.com/, http://www.deepmarket.com/, http://neinuclearnotes.blogspot.com/, and http://physics-envy.blogspot.com/ for the recent links.

 

Slurrier blogs GECCO

Bill Tozier (AKA The Notional Slurry, AKA The Slurrier) blogs about GECCO-2006 here.

 

Discovery of cis-regulatory elements with a genetic algorithm

here.

 

Discovery of cancer protocols with a genetic algorithm

here.

 

Knowing.net blogs the Humies

here.

Wednesday, July 19, 2006

 

Epictetus rides again

at TEE (here).

 

CEC part of WCCI in Vancouver

IEEE's big genetic algorithms conference (CEC) is going on now as part of the World Congress on Computational Intelligence in Vancouver, British Columbia (see here). WCCI always attracts a big crowd, and they have a large panel of plenary speakers in neural nets, fuzzy systems, and GAs.

Monday, July 17, 2006

 

MEDAL gathering on evolutionary computation

I am pleased to announce that MEDAL Gathering on Evolutionary Computation is going to take place next Monday, July/24/2006 at the University of Missouri in St. Louis. The primary purpose of this even is to celebrate the creation of the Missouri Estimation of Distribution Algorithms Laboratory (MEDAL), hear a number of cool presentations, discuss interesting challenges in research and applications of evolutionary computation and machine learning, and have a free lunch (yes, I've heard about the no-free-lunch theorem, too many times in fact). The presentations will include the following:

For more details about the event, please visit its web page. I'll try to provide the live-blogging service at the event, but I may just summarize it afterwards.

Friday, July 14, 2006

 

Journal of Natural Computing
Special Issue on Learning Classifier Systems

Guest Editors
Larry Bull, University of the West of England, U.K.
Pier Luca Lanzi, Politecnico di Milano, Italy

Call for Papers
Learning Classifier Systems were conceived almost 30 years ago by John Holland, the father of genetic algorithms. Since then a number of models have been developed to bring Holland's innovative ideas to real world applications. Submissions of original unpublished works are invited that investigate in different ways the state of research in Learning Classifier Systems.

Schedule:

Submission September, 30th, 2006
Notification December, 15th, 2006
Camera ready February, 15th, 2006


Submissions and enquiries should be sent to lanzi@elet.polimi.it.

Wednesday, July 12, 2006

 

Sastry, Lanzi, and Goldberg among paper winners

It was just announced that lab members Kumara Sastry, Pier Luca Lanzi, and Dave Goldberg were among the best paper winners at GECCO-2006.

 

APECS Conference in planning

Erik Goodman announced that planning is underway for a conference called the Asia-Pacific Evolutionary Computation Conference (APECS?). Plans are for a 400 person conference with 1/4 international attendance in October 2008.

 

Conference report 2006 and 2007

Mike Catallico (2006 chair) has asked for help in soliciting student travel grants from the government and industry. Student travel is dependent on these donations. 509 attendees came to the conference.

Hod Lipson (2007 chair) reports that the organization is moving GECCO to Europe (probably London, probably July 2007). There are big plans to have big name keynote speakers and an open Humie Exhibition.

 

Publications update

Wolfgang Banzhaf, editor of Genetic Programming and Evolvable Machines, reports that the journal is in its seventh year and going strong by recent measures of downloads.

Marc Schoenauer, editor of Evolutionary Computation, reports that MIT Press is looking at the subscription number reduction with some concern. Marc reports progress in reducing the delay between submission and publication.

 

SIGEVO bylaws

Erik Goodman is explaining that the executive board will be expanded from 15 to 18, that elections will be held every other year (from every year), and that the first election will elect 9 people to the board. Subsequently, 6 members will be elected at each election. Una-May O'Reilly is heading a nominating committee. A chair, vice-chair, secretary, and treasurer will be elected every two years by the executive board.

Erik has just surprised Pier Luca Lanzi by calling him to the podium. The 3rd issue will come out in September and will be devoted to a GECCO wrap-up.

 

Status of SIGEVO

Erik Goodman is reporting on the status of SIGEVO (www.acm.org/sigevo). In fall 2004, the ISGEC board approved transition to SIG status and on January 1, 2005, SIGEVO started life as a transitional SIG for its first year of operation. On June 15, 2006, ACM approved SIGEVO as a regular SIG.

There are currently 506 members, and we have now had two editions of our newsletter SIGEVOlution (Pier Luca Lanzi, editor and IlliGAL affilliate). 2006 marks the 21st year of genetic algorithms conferences (going back to the first ICGA in 1985 at MIT).

Erik just awarded a placque to Mike Cattalico (GECCO-2006 chair) and is thanking the cast of thousands that made GECCO possible.

 

2006 Humie winners announced

The Human Competitive Results competition (see here) results were announced at ACM SIGEVO's GECCO-2006.

Bronze:
Jie Yao, Nawwaf Kharma, Peter Grogono, A multi-population genetic algorithm for robust and fast ellipse detection
Gustavo Olague, Leonardo Trujillo, Using Evolution to Learn How to Perform Interest Point Detection

Silver:
Kumara Sastry, Duane D. Johnson, Alexis L. Thompson, David E. Goldberg, Todd J. Martinez, Jeff Leiding, Jane Owens
Multiobjective Genetic Algorithms for Multiscaling Excited-State Direct Dynamics in Photochemistry

Gold: Varun Aggarwal, Selçuk KILINÇ, Varun Jain, Uğur ÇAM, Catalogue of Variable Frequency and Single-Resistance-Controlled Oscillators Employing A Single Differential Difference Complementary Current Conveyor

 

Live blogging: SIGEVO biz meeting

It's the last day of GECCO-2006 here in Seattle and I'm blogging live from the ACM SIGEVO business meeting. Chairman Erik Goodman is at the podium and we should soon hear the wrap up of this year's conference, the get a prospectus for next year's conference, and learn the winners of the best papers awards and the Humies

Tuesday, July 11, 2006

 

Pier Luca on classifier prediction based on tile coding

Pier Luca Lanzi is just giving a cool talk Classifier Prediction based on Tile Coding on the best-paper LCS session.

 

GECCO-2006: Real-valued EDAs and EDA applications

I am now sitting in the EDA session that focuses on real-valued EDAs and EDA applications. A while ago Joern Grahl finished his talk on adaptive variance scaling in IDEAs. Right now, Yunpeng Cai is explaining how to use EDAs with Boltzmann selection in real-valued applications using Kullback-Leibler divergence in model building. The session will close with John McCall's talk on using EDAs for optimizing cancer chemotherapy.

 

Two talks in LCS and GBML @ GECCO 2006

Martin Butz is now talking on hyperellipsoidal conditions in XCS. Xavier is next who will talk about fast rule matching in LCS.

 

GECCO 2006: Representations in LCS

James Marshall presented and interested result about evolving multiple representations in XCS. Pier Luca Lanzi is now going a an interesting new condition scheme proposed by Stewart Wilson based on convex hulls.

 

Kalyan Deb on practical EC practiced in academia

I'm just attending Kalyan Deb's presentation Practical EC practiced in academia: How challenging has it been to me? Kalyan talks about challenges that academic researchers have to face when working on the real real-world problems with industry.

 

Anil Menon

Once genetic algorithm researcher Anil Menon is now a writer of fiction. See an blog interview with him at Duck of Destiny (here).

 

Of mice, genes, and brains

The 2nd keynote talk at GECCO-2006 is Mapping and Mining the Allen Brain Atlas by Dr. Mike Hawrylycz, Director of Informatics at the Allen Institute for Brain Science. Gene markers are used to map gene expression in the slices of mouse brains. See www.brain-map.org for more information.

Monday, July 10, 2006

 

Jason Daida on symmetry breaking in GP

I finished all my presentations today and now I can go attend presentations of many interesting papers this GECCO. Right now Jason Daida is presenting an interesting work on characterizing the dynamics of symmetry breaking in GP. Darrell Whitley is next and is going to talk about alternative evolutionary algorithms for evolving programs

 

GECCO 2006: Pier Luca Lanzi and Kumara Sastry open fire

Sometimes is sunny, sometimes is rainy. Unfortunately, Pier Luca and Kumara are presenting right now at the same time, and I cannot split in two. Pier Luca is going over an interesting work on computed predictions for XCS. Kumara is presenting his work “Multiobjective genetic algorithms for multiscaling excited state direct dynamics in photochemistry”; the paper is nominated for best paper on the real-world application track, and is a must-read one.

 

Evolvable artificial hand

This morning, I attended the session of evolvable hardware applications. One of the applications that most interested me was an evolvable artificial hand. It reads the nerve signals from one's arm and evolves the actions that one desires (like grabing, holding, etc). I think it's really cool and could be extremely useful for people who need prosthetic hands. There are also many other interesting projects going on in Higuchi's group. You can check them out here.

Sunday, July 09, 2006

 

IWLCS 2006: In the heat of the discussion

We are having a discussion session about what, where, and when next? The question that sparked the fire: "Is reinforcement learning the enemy?".

 

Xavier on the Extended Compact Classifier System

Xavier Llora is just giving a talk on the Chi-ary Extended Compact Classifier System at the LCS workshop, where he is showing how we can incorporate linkage learning into the Pittsburgh learning classifier system using the ideas from estimation of distribution algorithms (EDAs), specifically, Georges Harik's extended compact GA. Important and interesting work.

 

I think therefore I am (an engineer)

Descartes for engineers (here).

 

Risto Miikkulainen's tutorial Evolving Neural Networks

A while ago I came back from the Evolving Neural Networks tutorial by Risto Miikkulainen, who's done a great job at presenting important concepts, results and challenges in neuroevolution. The tutorial also included a lot of impressive demos.

 

Learning classifier systems for bioinformatics @ IWLCS 2006

Jaume Bacardit is talking about ensemble techniques for Pittsburgh learning classifier systems, specifically applied to learning and classification problems in bioinformatics domain.

 

IWLCS and the Soccer World Cup Final

Planning is useful. Planning sometimes works. Others, it is just a guarantee that things won’t happen that way. IWLCS timetable is one of these won’t-happen cases. The presentations are being dynamically reordered based on people's need to watch the Soccer World Cup final.

 

Pier Luca on computed predictions

Pier Luca Lanzi is giving a nice talk about computed predictions in classisifer systems. at the LCS workshop.

 

Only a game

I'm sitting in the LCS workshop at GECCO-2006 in the Renaissance Seattle Hotel in beautiful downtown Seattle. Xavier Llora is announcing program changes to allow Pier Luca Lanzi and other Italians to watch the World Cup finals between France and Italy.

Saturday, July 08, 2006

 

Tutorial on probabilistic model building @ GECCO 2006

Martin Pelikan is giving a tutorial on probabilistic model building genetic algorithms or estimation of distribution algorithms. The tutorial slides can be found here.

 

GECCO 2006: First pleasent surprise

I had my first pleasant surprise today at GECCO. I am sitting in the Workshop on Medical Applications of Genetic and Evolutionary Computation (MedGEC 2006). The workshop contains up to four different papers related to LCS and other GBML topics for medical domains. This is 4 out 6! :D

 

Live from Seattle its GECCO

I am currently attending Franz Rothlauf's tutorial on representations in genetic and evolutionary algorithms.

Friday, July 07, 2006

 

Has engineering lost its mojo?

See TEE here.

 

GECCO is less than a day away

Xavier, Tian-Li, and I are at Champaign airport on our way to Seattle to attend GECCO-2006.

 

SIGEVOlution in the ACM Digital Library

The first issue of SIGEVOlution, published in April 2006, is now available from the ACM Digital Library and can be accessed from here .

Thursday, July 06, 2006

 

LCS and other GBML warming up for GECCO 2006

The agenda for the Ninth International Workshop on Learning Classifier Systems (IWLCS’2006) can be found here. The workshop is coming with a list of very interesting papers and topics. We are looking forward to another edition of the workshop crowded of new and exiting ideas. If you are in GECCO, do not let it pass by ;).

By the way, if you want to did a little further into the learning classifier systems (LCS) and other genetics-based machine learning (GBML) world, check the website of the NCSA/IlliGAL Gathering on Evolutionary Learning (NIGEL’2006). It contains the slides and videos of the talks on cutting-edge LCS and GBML research. It is just fresh out of the oven!

Wednesday, July 05, 2006

 

SIGEVOlution Volume 1, Issue 2

The new issue of SIGEVOlution, the newsletter of SIGEVO, is now available for you to download.

The June 2006 issue features:

The PDF can be found here. The newsletter is intended to be viewed electronically.

Monday, July 03, 2006

 

A philosophical summer

Read about my philosophical summer of reading and viewing at TEE here.

 

3 IlliGAL Members or Alums vying for 2006 Humie

Koza's Human Competitive Machine Intelligence Competition (The Humies) has 8 applicants vying for $10,000 in prize money (see here). Two of the entries have ties to IlliGAL. Kalyan Deb's entry (with Aravind Srinivasan) is entitled Innovization: Innovating design principles through optimization, and Kumara Sastry heads up a 7-person UIUC team concerned with Multiobjective Genetic Algorithms for Multiscaling Excited-State Direct Dynamics in Photochemistry. Best of luck to all contestants, but the judges need to understand that I have a kid going to Northwestern University next year.

 

SIGEVO a real ACM SIG

One of the many things to be celebrated at this year's GECCO is SIGEVO receiving designation as a full-fledged SIG. The status over the last year has been as a provisional SIG, but with the publication of the newsletter, and the holding of two GECCO's under ACM supervision, the training wheels have come off.

 

Less than a week until GECCO in Seattle

News from GECCO command is good. Registration is strong, and Saturday begins what promises to be a terrific conference. GECCO-2007 is tentatively set for July in London, the first time GECCO will leave the US. We like our lizard versus the insurance company's anyday (compare here and here)

This page is powered by Blogger. Isn't yours?