IlliGAL Blogging

Here's a job posting for a postdoc at the Free University, Amsterdam, to use genetic algorithms to do environmental modeling. The object is to "generate lessons for policy aimed at promoting and steering transitions."

Martyn Amos's blog has a post on his co-authored paper on Second Generation Biocomputing here. An abstract of the paper may be accessed here, but the following excerpt catches the gist of it:

Previous solutions (the "first generation" of biocomputing techniques), whilst reasonably effective, are crude analogues of actual biological systems. We believe that a new, inherently inter-disciplinary approach is needed for the development of the emerging "second generation" of bio-inspired methods.

There are a number of ways to object to this. We are arguably a good bit beyond the first generation of biologically inspired techniques already, others have previously called for more biology in biologically inspired techniques, and arguably the greatest advances in speed and power have most recently come from dropping fidelity to biology (think EDAs and model builders, for example). I agree that an interdisciplinary approach is needed to advance biologically inspired computing, but it is a greater interdisciplinarity than imagined by the authors of this position paper. In my view, advances in technique come from an interdisciplinary process involving (1) artificially and naturally inspired mechanism, both, (2) effective theory (in the form of little models), and (3) careful implementation and bounding experimentation. Calling for more biology is just one part of this and it is not clear that it is the most important part.

Dan Ashlock has a new textbook on evolutionary computation (see here).

TEE says "maybe" and has a theory here (with the help of Victor Davis Hanson).

Machine Learning, Etc. has a nice analysis of the time history of a number of machine learning related keywords using Google scholar here. The data for "genetic algorithms" would seem to suggest that interest in GAs has peaked or leveled-off. Is this so, or is the work in "genetic algorithms" now diversifying into a number of other names (genetic programming, EDAs, co-evolution, evolvable hardware, etc.)?

Papers submissions at the GECCO have been largely level since 1999 with an uptick or two over the last two years. My own view is that the field has been taking a breather and has been consolidating the gains of the 90s. As recent progress in competent GAs and supermultiplicative efficiency speedups takes off, I predict that interest will continue at current levels or possibly grow again. Growth will be particularly robust if current work in competent GP can be combined with supermultiplicative efficiency enhancement. (via KaraNagai)

Flags and lollipops says so here.

I've posted the table of contents of chapter 5, "Write for Your Life," of my forthcoming book, The Entrepreneurial Engineer, over at the eponymous blog (see here). The chapter is divided along process and structural lines with the first part of the chapter devoted to learning to separate creation from criticism and the second part of the chapter devoted to fundamental structure of biztech writing, B-P-R or background, purpose, and roadmap. It is my contention that all business-technical writing shares a common iteratively hierarchical structure and that once this structure is mastered most biztech writing becomes a snap.

Genetic argonaut has a post on the history of evolution strategies here. It is quite a nice post, although my comment tries to set the record straight on both historical and editorial counts:

Nice post on the history of EC. The first "official" meeting of German and US EC interests was actually PPSN in Dortmund in 1990, not ICGA 91, and PPSN was predated by my trip to a German OR conferernce Ulm in 1989, I believe, at H-P Schwefel's invitation. Yuval Davidor and I were at that conference representing Michigan-style GAs. Also, I'm not sure what the (unbalanced) editorializing about building blocks is doing in an otherwise very nice historical post. At the time, we knew there were differences of opinion about what was important, but we were all glad to have newfound comrads in EC arms. An accurate historical portrayal should emphasize the coming together that started in the late 80s and 90s.

Nonetheless, the post is a good one. We can hope that the argonaut does an equally nice job about the early days of Fogel and Holland.

Two students at IIT Madras have used genetic algorithms to optimize the bowling motion in cricket (see here).

See the relevant post at The Entrepreneurial Engineer.

This sounds interesting (via niche).

I'm spreading the gospel of integrated model building at the Frontiers of Science Symposium at Nagoya University in Japan today and tomorrow. Click for the conference home page and program. My talk title is Solving Larger, Faster, and Harder: The Incredible Story of Supermultiplicative Speedups. I'm one of the few soft computing guys here, but it's fun to see where hard computing has gotten to over the years since I got out of the business.

Doyouwantcoffee (an architecture blog) has a post (here) about a company in Singapore that is using genetic algorithms in product design. The company, Genometri (I like the name), has a variety of tools, but the ones of relevance to GAmists are Gennovate and Design DJ. According to the site, Gennovate is

Entry Level generative software. Creates hundreds of variations based on a generic model built in Solid Works. Most of the generative designs seen in this web page were generated using gennovate. Status: currently being alpha tested.

and Design DJ

Will give you ultimate control over the design generation process, by allowing users to set limits on variations and allowing them to save the part file in a genetic format. Status: under-development.

The company won a business plan competition in April 2005 (see here) and is apparently a spinoff of research at NUS. Jasmeet Chadha and Parth Vaishnav are the principals listed on the b-plan competition, but there are no names explicitly mentioned on the Genometri site about company officers, investors, and such. From a technology point of view, there also isn't much to go on, but the cool stuff appears to be the integration with CAD tools, not particularly advanced genetic algorithms.

I'm in Narita airport in Japan on the way to Sydney, Australia for ACAL 2005 and AI 2005. Giving a talk called Little Models, Big Results (see here).

The Journal of New Mathematics and Natural Computation is running the first ot two parts of a special issue on chance discovery (volume 1, number 3). A quick note and links to it may be found here.

Mister Snitch has a nice post about the different styles of bloggers that make up the ecology of blogging here (via Instapundit). Related posts are here and here.