MEDAL Blogging

Joaquín M López Muñoz discusses the link between human innovation and evolutionary computation based on selection, crossover and mutation. He also argues that there are lessons to be learned from evolutionary algorithms to improve the innovation process.

David E. Goldberg discusses this in numerous works, for example in his Design of Innovation book. The second edition of this book, Genetic Algorithms: The Design of Innovation, coauthored by Kumara Sastry, should hopefully soon be out.

GreyThumb pointed out an article Lots of Animals Learn, but Smarter Isn’t Better published in New York Times.

The article argues that flies can evolve the ability to learn relatively fast, but being able to learn fast may reduce their survival abilities. This observation is used to support the argument that being smarter (in the sense of being able to learn faster) is not always a great thing.

The article reminded me of the movie Idiocracy but that’s another story :-)

The photo courtesy of National Geographic.

Ben Goertzel published an interesting post Open-Source Robots + Robot Simulators + Virtual Worlds + AI = ??? on his blog The Multiverse According to Ben.

Mark Hauschild just gave a great talk Using Previous Models to Bias Structural Learning in the Hierarchical BOA as part of our departmental colloquium series. Mark argued that EDAs provide us with a lot of knowledge about the problem in the form of probabilistic models learned from the populations of promising solutions and that throwing this information out is a waste. Instead, he suggested that we use those probabilistic models from past EDA runs to speed up future runs on similar problems. The special focus was on the hierarchical BOA (hBOA).

I think that using prior problem-specific knowledge and learning from past runs to speed up future ones are among the most important lines of research in EDAs. Some of the material on this topic can be found in MEDAL Reports No. 2008003 and 2008007.

We just put up MEDAL Report No. 2008007, Enhancing Efficiency of Hierarchical BOA via Distance-Based Model Restrictions by Mark Hauschild and Martin Pelikan. The abstract follows:

This paper analyzes the effects of restricting probabilistic models in the hierarchical Bayesian optimization algorithm (hBOA) by defining a distance metric over variables and disallowing dependencies between variables at distances greater than a given threshold. We argue that by using prior problem-specific knowledge, it is often possible to develop a distance metric that closely corresponds to the strength of interactions between variables. This distance metric can then be used to speed up model building in hBOA. Three test problems are considered: 3D Ising spin glasses, random additively decomposable problems, and the minimum vertex cover.

SiouxWIRE Annex published an interview with Theo Jansen on his creation, the autonomous, wind-powered machines. The machines are composed mainly of plastic tubes and lemonade bottles. A genetic algorithm was used to determine a proper combination of tube lengths to properly translate circular movement to movement of the machine “toes”. Several related videos can be found on YouTube:

I was just preparing a lecture on particle swarm optimization and ant colonies, and while searching the web for materials on the topic, I found the following nice video demonstrating how ants search for shortest paths. Check it out.

The deadline for the 5th annual HUMIES Awards (Human Competitive Results Produced by Genetic and Evolutionary Computation) is May 26, 2008. HUMIES is organized within the GECCO-2008 conference. For more information, visit HUMIES 2008 web page.

The website of Computational Learning Theory (COLT) and the association for Computational Learning (ACL) has been remodeled. Check the updated web page here. Don’t miss the interview with Vladimir Vapnik, which is provided as part of the planned series of interviews with COLT leading researchers.

A new book entitled A Field Guide to Genetic Programming by Riccardo Poli, William B. Langdon, and Nicholas Freitag McPhee (with contributions from John R. Koza) has recently been published. The PDF of the book can be downloaded for free here or a cheap printout of the book can be purchased through lulu.com here. See related posts at IlliGAL Blogging and Michael Trick’s OR Blog.