IlliGAL Blogging

I just want to drop a quick note. We finally re-vamped the LCS page and turned it into a blog :). We are planning to use it initially to mainly post the news about the activities of the International Workshop on Learning Classifier Systems. Maybe later we add some other LCS and GBML related info. We are doing our best not to overlap the usefulness ot the LCSWeb wiki that Alwyn Barry started and Jan Drugowitsch bring to its current form. The web will be soon reachable at http://www.learning-classifier-systems.org, but, while it is in beta stage, you can access to it here.

Waltham, Massachusetts company Affinnova has named Waleed Al-Atraqchi as chief executive officer (see full press release here).

Waleed is a proven leader, one with the optimal combination of sales and marketing management and entrepreneurial expertise needed at this stage of Affinnova's growth," said David Andonian. "His track record of elevating companies affords him keen insights into the critical success metrics needed to take our organization to the next level. He is a well-chosen addition to Affinnova's growing roster of consumer packaged goods and marketing services experts.

Affinnova applies interactive genetic algorithms to product design:

Affinnova harnesses the voice of consumers to help companies develop more successful brands and products. Its unique, patent-pending technology uses genetic algorithms to "evolve" product designs in response to consumer preferences. Affinnova's technology has been adopted by a growing list of blue-chip customers including eleven Fortune 500 companies. Affinnova is privately held and headquartered in Waltham, Massachusetts. For more information about Affinnova, visit www.affinnova.com.

An earlier IB post about Affinnova is available here.

Insolublog has an interesting critique of the modern university here. His criticism resonates with my blogging about seeking a college for my son Max. The basic point is to somehow inject mentorship back into the college or university experience:

Not everyone is a natural teacher, but there are many with natural gifts for creative building and innovation. Charlie was that rare blend of natural teacher and renegade technologist. He got all of that power from his own study and family dynamics. Apprenticeship is a old, tried and true concept, preceding all of the modern institutions of learning. Can this old concept be reinvigorated in our time, for more than the traditional electrical, plumbing and legacy trades?

Perhaps attempting to tap this mentor to apprentice communication will destroy it. Small team chemistry is meant to be small. Maybe it is best left alone.

Interestingly, large universities have this kind of chemistry. It's called graduate school, but unfortunately that kind of small-team chemistry is almost impossible on a large campus given the pressures toward funded research. Some smaller liberal arts colleges with 1200-2000 students have found this small-group chemistry, and that is why about half of the schools on Max's final list are of that type.

Over the last few months I've mixed my thoughts about engineering and professional development into the mix of normal genetic algorithm stuff here at IlliGAL Blogging. Starting today, I'll blog about such matters at the new blog The Entrepreneurial Engineer.

On Wednesday, I submitted the final manuscript for TEE the book to Wiley for publication by summer or fall of 2006. Come on over to the The Entrepreneurial Engineer and keep in touch with your inner engineer.

The post "Competent GAs revisited" (see here) has gotten a string of 8 (!) comments that I would like to comment on. The debate is over whether genetic and evolutionary computation will continue to be an artform, where the practitioner chooses the representation, "appropriate" crossover, mutation, and other operators, or whether competent GAs obviate the need for such choices.

To frame the question, let's ask what a "typical" GA/EC user does in "choosing" operators and codings. First, he chooses an initial representation and operators, gains "experience" with that choice, and guides the choice of subsequent representations and codings with the "lessons learned." The processes of choice and experimentation in this tableau are less than well specified, but essentially the user "data mines" the stream of examples tested to do better.

What does a "competent GA" do? It "mines" the data stream systematically and uses the probability distribution of adapted operators and the population to sample new points. More recent work, uses these structural models to build a fitness surrogate, cheaply and accurately, and new work is pointing in the direction of automatically adapting mutation operators and other local search techniques in concert with the competent GA.

Seen in this way, competent GAs systematize the informal experimentation of the practitioner and replace it with sound decision making according to what is actually learned about the current landscape. In this view, there is nothing special about the actions of the practitioner; it should be possible to replace the informal and unsystematic experimentation of the practitioner with machine-based techniques that do a better job on many if not most problems.

Thus, as competent GAs and their derivatives take hold, I predict that there will be a decrease in the numbers of bad GA/EC papers advocating weird combinations of operators and weird techniques on strange grounds and an increase in landscape learning and adaptation. I certainly hope so.

Geekgorgeous has a calendar of geek models. My favorite is Lilac:

Lilac, who started working as a programmer at age 16, is now a senior software engineer with an acronym-rich skill-set that includes Java, J2EE, EJB, JSP, JMS, PHP, ASP, ADO, SQL, XML, UML, J2ME, MIDP and more. As an independent consultant, Lilac has worked on numerous business applications on various platforms, from home-automation systems to portfolio optimizers based on genetic algorithms. She's been a technical session speaker at multiple conferences, including Sun's Java One conference, the Borland Technology Conference, and the Denver Java Users Group. A “numbers girl”, Lilac’s latest obsession is data mining and statistical analysis of econometric data for an upcoming real estate investment application. Lilac also owns three video-game stores in Colorado.

See Lilac and her 11 friends here. Hat tip technojunk.

I'm giving a workshop (announcement here) tomorrow morning (Friday) in Harrisburg called The Joy of Engineering in an Age of Opportunity. It ties into the themes in my forthcoming book The Entrepreneurial Engineer (here) and my shortcourse of the same name (see here).

This is not a typical blog entry. I felt like telling you before you continue reading. Last Saturday I was sitting in one of the dance performances of this season at Krannert center. I was amazed by the hard work done by Philip Johnston's students---soul lifting is not quite the word, but it is the closest I can think of. Anyway, I was sitting in one of the intermezzos, when a question in the audience struck me. “Is there any new notation for dance?”.

New? That meant that there were already dance notations. I just browse a little bit to ease my ignorance, and I run into several entries here. Moreover I found an online paper on dance notations and computers. After reading it, I cannot stop thinking about how the same DISCUS technology used for scenario creation in marketing could be applied to collaborative on-line concept discussions for dance choreographies. And they already have notations for it! I know, maybe it is my naiveness, but I think it would be useful for companies scattered across the globe to boost their interactiveness.

Blogger and regular reader Nosophorus asks the following:

...I only know that those Competent GAs "refer to principled procedures that solve a large category of hard problems quickly, reliably, and accurately". However, there is some canonical form for those GAs ?? There is some paper that reports about those GAs...

I'm glad you asked. Competent GAs take many different forms, but they all share certain characteristics and principles of operation. First, they are population based and they all possess an adaptive or self-adaptive crossfertilizing mechanism that learns and exchanges effective substructures. Second, the use of a population and crossfertilization mandates a certain physics of initialization, competition, decision, and exchange.

Interestingly, since 1993 and the fast messy GA, many different types of mechanisms have been developed for competent GAs, and if you would compare one to the other, you wouldn't recognize their similarities. For example, outwardly the fast messy GA looks nothing like hBOA. Only at the level of theory of operation do their similarities appear. There are many papers, theses, and dissertations that establish these principles going back to the late 80s and early 90s, but my book The Design of Innovation (see also here) is a one-stop source for surveying the principles and some of the mechanisms of competent GAs. The book provides the basics and you can dig into the different areas of concern at your leisure. A google searchable version of DOI is available here.

An earlier post by Martin Pelikan made some laudatory remarks about Marc Toussaint's compression evolutionary algorithm (see here), and on the face of it the work appears to rival the best competent GAs, but further investigation shows that the compression EA only performs well on problems with tight linkage. This is a severe limitation, and the assumptions of the work should have been more clearly stated.

I ran into an interesting article by Clay Shirky here. The article explores the blog ecosystem under the light of power laws and social networks. This is a snippet of the article:

In systems where many people are free to choose between many options, a small subset of the whole will get a disproportionate amount of traffic (or attention, or income), even if no members of the system actively work towards such an outcome. This has nothing to do with moral weakness, selling out, or any other psychological explanation. The very act of choosing, spread widely enough and freely enough, creates a power law distribution.

Some claims are intriguing.

IlliGAL Blogging readers will have noticed an increase in comment spam on the site, whereby irrelevant comments are posted by bots or other malicious entities. To fight this scourge, IB has taken two steps. First, the blog will no longer accept anonymous posts. All commentators will be required to have a Blogger account, and if you don't have one, they are easy enough to obtain. Second, commentators will need to authenticate their humanity by retyping a word shown on the screen in a hard-to-machine-read font. IB hopes that these two steps will reduce comment spam sufficiently for us to continue to permit comments on the site.

Judea Pearl stopped by UIUC today. He gave a talk entitled “The mathematics of causal reasoning” at Siebel Center. I won’t highlight the contents of the talk, I won’t be able to make justice to it. In a nutshell, it was a blast. However, I want to post a sentence of his lecture:

Computer science is the science of wishful thinking.

People laughed, but I am not sure how many people got truth hidden behind what seemed a casual joke. He mentioned computer science, but I guess that it is also extensible to any other human endeavor involving any degree of innovation and creativity; innovation not afraid to transgress current conventions (scientific, social, or cultural), and creativity that dreams about a vision that may, eventually, become reality.

These words sounded pretty powerful to me, when pronounced at the home of ILLIAC, and deserve more than a casual laugh. The truth is that Judea’s talk went beyond the content. It was a talk about how not to dismiss wishful thinking, because it is the fuel that powers human innovation and creativity.

Check out the new DISCUS (distributed innovation and scalable collaboration) site here. A previous story about DISCUS is available here. One cool feature is that you can run your own KeyGraph by pasting in appropriate text.

Over the weekend, Max and I visited Eckerd College in St Petersburg, FL as part of our continuing series of college visits (see here to traceback earlier blog entries). The setting was lovely, and the food and dorms were as nice as any we have seen. Compared to the other small schools we visited (Earlham, Kalamazoo, Ursinus), Eckerd is less academically intense, and the drug and alcohol policies seemed fairly lax to the point of being almost nonexistent. Nonetheless, the unique marine science program, extensive international travel, internships, and faculty-student interaction were impressive.

My new short course, The Entrepreneurial Engineer, is online and ready to go here. The online course uses video and synchronized powerpoint in eleven lecture modules (see here) to cover key personal, interpersonal, and organizational skills necessary for engineering success in an age of opportunity. Check out the free course preview here or signup for the full course online here.

I recently signed a contract with John Wiley and Sons to publish the book upon which the short course is built.

In workshops, it is a commonplace to take a large group meeting and divide it into a number of equally sized subgroups to give each person more of a chance to participate in the same amount of time. To share the discussion results at the end, each group presents a short report to the full assemblage.

In a recent IlliGAL technical report, A Little Model of Optimal Group Size in Breakout Meetings, I present a little model of the tradeoff between reporting and discussion in breakout groups and calculate the optimal group size. The note (pdf) is short, but the result is sensible, and it ties to a number of other useful little models in organizational settings. Tian-Li blogs about his presentation at the Academy of Management (here), and our note on optimal team size under deciding and doing is here, and our optimal communication in a hierarchical organization paper is here.