Friday, September 30, 2005

 

The social network of evolutionary computation

Carlos Cotta just wrote me an interesting e-mai:
I'm contacting you regarding an analysis we have recently done on the structure of the evolutionary computation community using social network techniques. Besides macroscopic properties of the collaboration network, we have studied which the central actors (that is, authors of EC papers) of the network are on the basis of several centrality measures.
A pdf file of the study by Cotta and Merelo is available here. Although yours truly is numbered among the central figures in the field, Kalyanmoy Deb tops the rankings in terms of the traditional SN measures of number of co-workers, betweeness, and closeness.

Thursday, September 29, 2005

 

Still programming in Fortran?

Then this site reminded me of Dave Carroll's fortran genetic algorithm code (see here). Its a fairly simple GA with more than the usual bells and whistles.

 

ArchiKluge is right

ArchiKluge is a pretty java-based demo that

ArchiKluge is a simple Genetic Algorithm that evolves architectural diagrams. It explores the qualities of design made by machines, devoid of any intention, assumptions or prejudices, and which often display a very peculiar form of mindlessly but relentlessly pounding against obstacles and problems until overcoming them, a manner of acting nature and machines commonly exhibit.

Well I'm not sure about any of that, but the design of the user interface is pretty cool. The actual "problem" being solved is not very interesting, and it isn't clear what the GA is really learning, but pretty pictures will trump substantial inquiry every time. Hat tip blprnt.blg.

 

Hybrid GAs

Hybrid Car Reviews has a quirky collection of articles on hybrid genetic algorithms here. When doing theoretical studies, it is important not to make matters messy by throwing to many helpers into the mix. In applications studies, most practitioners find it useful to hybrid GAs with some local procedure.

Saturday, September 24, 2005

 

The modeling spectrum and entrepreneurship

For the past year, by day I've been a mild-mannered engineering professor, but by evening I've been a wild and crazy chief scientist for a start up in ubiquitous personalization called Nextumi (see here). The other day I was thinking about our progression from the formation of the company to our current state and how that journey represented a trip along the modeling spectrum (see here, here and here, ) in terms of the case for the company.

At first, our thinking about the company was driven by intuition and barely articulable thoughts. As we planned more, we were able to articulate an increasingly elaborate story of what the company did. As time progressed, we got more evidence for the correctness (or incorrectness) of that model, and we adapted and moved on; at some point we started to make financial projections of what the company might earn, and we shifted from a qualitative view to a more quantitative perspective.

It is particularly surprising how powerful and important the early stories have been to gaining interest, investment capital, and indeed driving the development. Every company has a creation story, and stories of perils and heroism along the way, and in many ways these stories are as important to understanding and valuing a young company as the P&L and balance sheet.

Wednesday, September 21, 2005

 

The 1st Workshop on Evolutionary Computation in Italy, finally!

Italy is a small country. Artificial Intelligence in Italy is a very small community, mainly consisting of researchers in the so called "classical AI". However, yesterday the 1st italian workshop on Evolutionary Computation took place during the 9th Conference of the Italian Association for Artificial Intelligence. Around 20 researchers and other 20 master students of the University of Milan gathered in a room and discussed their work. Some of them are very well-known (Stefano Cagnoni, Riccardo Poli, and Marco Tomassini) but I also met people who use EC methods and do not publish neither in GECCO nor in any other EC conference.

Among the others, I found Riccardo's tutorial on "Doing research in evolutionary computation" very entertaining. I also liked Marco's tutorial on "Spatial and temporal dimensions in evolutionary systems". Unfortunately I had to leave before the final round table. But they told me that in 2006 they plan to organize the 2nd one!

Monday, September 19, 2005

 

The Fighting Illiacs?

The University of Illinois has drawn heat from the NCAA and its recent ruling on mascots. The Fighting Illini are represented at halftime shows by a fellow wearing Indian headgear and doing a dance, and campus has been exercised about this for some time (see here). I'm something of an agnostic on the issue, and it seems too bad to give in to the forces of political correctness, especially when they seem less exercised over mascots that might be offensive to others (Fighting Irish, USC Trojans, MSU Spartans, etc.), and what about all those animal mascots. Hasn't anyone ever heard of PETA?

Regardless, in the interest of promoting campus harmony, I have had a brainstorm that should settle matters once and for all. The campus should pay homage to its own role in the creation of the modern computer & IT eras and rename the team either The Fighting Illiacs or The Bucking Browsers. The first recalls the first computer built and owned by a university (see here) and the second recalls the invention of the Mosaic browser at Illinois at NCSA.

A side effect of the choice is that it will sidestep the banishment of animal mascots by the NCAA in 2067 at the behest of animal-rights groups.

Saturday, September 17, 2005

 

The joy of engineering

Chapter 2, entitled The Joy of Engineering, in my forthcoming book The Entrepreneurial Engineer (see previous posts here and here) covers a variety of topics of interest to engineers and computer scientists.
Chapter 2 The Joy of Engineering
2.1 A Joyous Confession
2.2 Engineering as Liberal Education, Launch Pad & Lifelong Love
2.2.1 Who is getting the real liberal education?
2.2.2 Engineering as launchpad
2.2.3 10 ways to love engineering
2.3 The Fundamental Tug-of-War
2.4 Science and its Little Secret
2.5 Engineers: First Masters of Modern Enterprise
2.6 Economy of Intellection: Separating Science from Engineering
2.7 Four Tensions Facing the Entrepreneurial Engineer
Summary
Exercises

The online short course will be available later this fall (here) and the book should be available in 2006.

 

Electric sheep screen saver uses distributed interactive GA

The Program Witch Pages has an interesting post on a screen saver called Electric Sheep that evolves abstract images (called sheep) using a distributed interactive genetic algorithm in which the images are evolved through the decisions of many screen saver users. The screensaver may be downloaded here. The program was conceived by CMU PhD grad Scott Draves (see also here).

 

New blog covers computationalism

A month-old blog called Computationalism presents a variety of reviews of books and articles of interest to genetic algorithmists and artificial evolutionaries. For example, Chris Langton's 1997 book collection of papers on artificial life is reviewed here. Gary Cziko's book, Without Miracles, is reviewed as well (see here). This interesting blog has been around since the beginning of August (it sneaked onto Blogger while we were all on vacation). The author does not identify him or herself, but the first post may reveal the author. It reviews a book by Li Jianhui of Beijing Normal University called Towards Computationalism: An Introduction to Philosophy of Artificial Life, apparently published in Chinese. Computationalism is defined (google define link here)
The notion that the operation of the mind can be explained entirely in terms of the formal, or functional, properties of a computational system. See also Cognitivism, Eliminativism, Functionalism, Materialism, Turing Test.

IlliGAL Blogging readers look forward to more interesting posts at the intersection of philosophy and computation from Computationalism.

Friday, September 16, 2005

 

Google blows blog search

Google now has a blog search engine in beta here, but it is surprisingly inept. Unlike Technorati, Google does not appear to use recency as a key element in its sort, so the results are all over the place. Try it here on "genetic algorithms" and see what I'm talking about. Compare it to Technorati on the same search here. I don't know about you, but when I do a blog search, I want to know first what's happening now, not what happened in March

Search on a URL isn't much better. Consider the Google blogsearch for IlliGAL Blogging here and compare it to the comparable Technorati search (here). It surprising that Google wouldn't even back mine Blogger, the blog publication site it owns.

Don't count Google out, but for now don't change away from your favorite blog search engines. Google's offering isn't what bloggers or blogreaders want or need.

Monday, September 12, 2005

 

Live from Genoa, Italy: ICIP 2005

I'm here in Genoa, Italy (tough duty if you can get it) attending the IEEE ICIP 2005 conference. ICIP is quite a large conference surveying the latest in image processing. See here for program details. I'll be giving a keynote about competent GAs on Wednesday morning and back to scenic Champaign (sigh) to teach class and work on a cool senior design project that evening.

Sunday, September 11, 2005

 

Deb wins Bhatnagar award

IlliGAL alum Kalyanmoy Deb has been named a 2005 winner of the prestigious Shanti Swarup Bhatnagar prize in Engineering Sciences. See here for more information. Kalyan has done more than any single person to spread the word on multiobjective GAs throughout the planet and he has done more than any single person to spread the practical application of GAs in his home country. Please join me in congratulating Kalyan on this singular honor.

Saturday, September 10, 2005

 

NSF, fingerprints, and GAs

The NSF has noticed that genetic algorithms are starting to rival human innovation and inventiveness. Hummie winning work out of the University of Texas at Austin received a nice writeup here. Other hummie blog entries may be found here and here. (via Duane Johnson)

Friday, September 09, 2005

 

IlliGAL paper in Physcial Review B & Virtual Journal

A recently accepted IlliGAL paper (Sastry, K. Johnson, D. D., Goldberg, D. E., & Bellon, P. (2005) Genetic programming for multitimescale modeling, Physical Review B 72, 085438) has also been selected for posting in the 29 August 2005 edition of the Virtual Journal of Nanoscale Science and Technology. The paper uses GP to bridge slow yet accurate materials calculations (MD or molecular dynamics) at the molecular level with fast statistical procedures (KMC or kinetic Monte Carlo). The tuned statistical procedures can accurately model systems with small numbers of molecular runs, thereby allowing accurate materials modeling 3 to 9 orders of magnitude faster than is possible with MD calculations alone. IlliGAL PhD student Kumara Sastry is jointly advised by Duane Johnson in Material Science and Engineering at UIUC and me.

Thursday, September 08, 2005

 

IEC at FSS2005

Six interactive evolutionary computation (IEC) papers were presented at an annual Japanese domestic symposium of SOFT held in Tokyo, 9/07-09, 2005.

A. IEC applications

A1: "Interactive Logo Drawing System"
Designing logo marks using IEC
A2: "A System for Beautifying Face Images Using Interactive Evolutionary Computing"
Tuning parameter of image processing system using IEC
A3: "An Interior Layout Support System with Interactive Evolutionary Computation using Evaluating Agents"
Designing interior layout with IEC
A4: "Behavior Acquisition of Four Legs Robot Using Interactive GAs"
Acquiring the preferable behavior motions of AIBO robot with IEC

B. IEC interface research

B1: "Interactive Evolutionary Computation with Evaluation Characteristics of Multi-IEC users"
Combining IEC and normal EC with prepared evaluation characteristics of IEC users that are used to select better individuals through simulation with big population size and accelerate EC convergence. Idea is to use evaluation characteristics of other IEC users till the user's evaluation characteristics are learned in early generations.
B2: "Subjective evaluation on the Method for Reduction of IEC user's fatigue through Rating-Scale Mapping"
Changing relative fitness values, which are frequently used in normal IEC, to absolute ones to learn IEC user's evaluation characteristics effectively. The leaned evaluation characteristics are used to accelerate IEC convergence and reduce IEC user's fatigue problem.

Wednesday, September 07, 2005

 

Highlights and Lowlights from CEC 2005

I just returned from CEC 2005 in lovely Edinburgh. I'll be honest in saying that I didn't get to attend much of the conference. However, I did see a couple of things I thought were worth sharing.

First, a highlight. David Wolpert gave a keynote on the 10th anniversary of the No Free Lunch theorem. It was very gratifying to see that Mr. Wolpert has no confusion about what's most interesting about NFL: the cases where it does not apply. In particular, he made some interesting comments on systems where you have biases about which problems are important, and systems where problems and solvers are co-evolved (Stuart Kaufman would be most gratified). In both cases, as one would expect, NFL need not apply. Wolpert also commented on how NFL is now being used as a defense of Intelligent Design. He shares my bemused disgust.

Second, a lowlight. I presented a poster (almost everything at CEC was a poster this time) entitled An Iterative Mutual Information Histogram Technique for Linkage Learning in Evolutionary Algorithms. The most common comment was "so, what's linkage?"

 

TOC for TEE

Here is a chapter-level table of contents for the forthcoming book, The Entrepreneurial Engineer:
  1. The Entrepreneurial Engineer: Ready for the 21st Century
  2. The Joy of Engineering
  3. Money, Work, and You
  4. Getting Organized and Finding Time
  5. Write for Your Life
  6. Present, Don’t Speak
  7. The Human Side of Engineering
  8. Ethics in Matters Small, Large, and Engineering
  9. Pervasive Teamwork
  10. Organizations and Leadership
  11. Assessing Technology Opportunities

Stay tuned for further revelations.


 

Common criticisms of GAs

are usually ill-informed. The blogosphere provides regular examples of this sort of drivel, for example here and here. The first is a rant about how general purpose solvers are doomed and the second complains how machine invention is something of a Rube Goldberg machine.

If blind optimization is as inadequate as NFLers insist, how did nature evolve the incredible complexity surrounding us? And if genetic algorithms and genetic programming are so weird, how come these procedures are regularly infringing on patents of human inventors and creating new patentable gizmos?

Many of these complainants seem stuck in a time warp of GA research as it was ten years ago or so (20?). Please wake up and smell the coffee. The field of genetic and evolutionary computation is roasting hot beans, grinding them just right, and brewing a lovely cup of Joe. Instead of griping about some perceived difficulty within the bowels of our percolator, you might want to take a sip, and learn why this field, not only continues to survive, but insists on thriving, flourishing, and breathing life into research across the spectrum of human endeavor.

 

GAs used in protein structure problem

Researchers at the University of Southhampton report success using genetic algorithms to evolve hidden Markov models for the prediction of secondary structure in proteins.
New results are presented for the prediction of secondary structure information for protein sequences using Hidden Markov Models (HMMs) evolved using a Genetic Algorithm (GA). We achieved a Q3 measure of 75% using one of the most stringent data set ever used for protein secondary structure prediction. Our results beat the best hand-designed HMM currently available and are comparable to the best known techniques for this problem.
An abstract of the work is available here.

Thursday, September 01, 2005

 

Jobs on jobs

One of the themes of The Entrepreneurial Engineer (see earlier post) is to find engaging work. My colleague Ali Abbas passed along a link to Steve Jobs's 2005 commencement address at Stanford (here). Its short, moving, and spot on. It isn't easy to find what you love and do it for a living, but those of who get to do so should be more thankful than we usually are.

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