Genetic Algorithms, Complex Systems, Economics. . .

Interesting on-line paper this one. It shows how the Santa Fé Institute's researchers, genetic algorithms, and complex systems rediscovered a classical way of economics -- but, of course, there is more than just that on the paper. It's worth reading!

Have you ever wondered how free markets could be self-adaptable?

Evolutionary Computation Researcher In Antarctica Has Found The Best Time Complexity Evolutionary Algorithm

My news webservice has brought to me the following evolutionary computation findings:

"An international mission based in Antarctica in a special lab has found the best time complexity evolutionary algorithm ever! Professor Pangloss Van Thunder-ten-Tronckh is the director of the project, which began along the late 1970s and had the former U.S.S.R. as the main manager beside U.S.A - Professor Trollov Buraninev is the director of the U.S.S.R. part. 'It is a finding that will open the doors of a new kind of science among all other sciences', said Professor Pangloss. He continues: 'This discovery is to evolutionary computation as the calculus was to physics and mathematics.'

Professor Pangloss states that his team was the first to solve an over a billion variables problems and they did it using just an AMIGA 500. He comments: 'Although some guys from Urbana-Champaign have claimed that they are the first ones, it is not true! As we are funded by CIA and the former KGB, our results never made the big mainstream publication magazines!'.

He claims he was the inventor of the Estimation of Distribution Algorithms (EDA): 'What outside Antarctica evolutionary computation researchers call EDA, was invented by me when I was an undergrad student during the 1950s! It was a homework that my advisor gave me.'

He also said that what Professor Hans-Georg Beyer has made upon the theory of evolution strategies is, indeed, a less elegant version of what Professor Buraninev's team had already done in 1981. Professor Buraninev states: 'I know very well the work of those German guys! Of course, it is just a lower quality flavour of my own research that I started in April 1st, 1977. What Professor Beyer has made is valid only under certain conditions, that is, Gaussian mutation, intermediate recombination, and comma/truncation selection - let alone that my self-adaptation mechanism is much more general than Schwefel's and/or Rechenberg's, since it can adapt to any kind of fitness/objective function! We made a complete theory of evolution strategies under all imaginable conditions, taking into account correlations, covariances, discrete recombination, noise, plus selection, comma selection, Gaussian mutation, Cauchy mutation, Gamma mutation, Geometric mutation, Bernoulli mutation, Poison mutation, Laplace mutation, uniform mutation, Zipf-Mandelbrot mutation, Student mutation, and so on... both univariate and multivariate distributions! And I did it in my special built Agat computer! And... remember: In Soviet Russia genetic algorithms rotate... YOU!!!'

Professor Pangloss comments about a strange guy that once entered his iglu: 'Yes... he was very strange! He stayed here for almost two weeks and he worn every single day only striped shirts! He promised me that he would teach us how to always win when betting in lottery. I thought him a nice guy and showed him my new invention - Programs that generates, through simulated evolution, others programs! I gave him a source code in LISP! To my very own desperation, years later I was told that he had invented an invention machine and was using my code on it!'

Professor Buraninev once met a Californian guy from La Jolla: 'Yes! That guy once told me that he would introduce me to a friend of his, her name was Blondie24. At a first glance, I thought her name a little strange, since girls do not use to put numbers on their names. But he stated that that was a new Western fad and that the girl was a true master in checkers. Oh, heavens! I would finally realize my Soviet Nerd dream: Date a girl which also is a checkers master! Then, he fixed a day for a meeting. I believed... how fool I was! While the girl, the so-called Blondie24, had not arrived, he showed me some of his father's work on automata. I decided to show him my new evolutionary algorithm flavour which is almost the same as evolution strategies, except on the selection mechanism and the self-adaptation procedure. The guy went insane! He immediately ran away from the Iglu Cafe! Some years later, I got aware that he was engaged in a new business endeavour and using my mind child!! The girl, Blondie24, has never appeared!'

Well, despite all those confusions along the years, Professor Pangloss is really happy to announce the ultimate finding of his lab team. He says: 'Finally, after 30 years of research, we are able to declare that we have found an evolutionary algorithm whose complexity is only O(n^-100-100)! This is the complexity of our algorithm in ALL classes of problems! Both decomposable - the easy ones! - and non-decomposable!!' So, where is such a poweful evolutionary algorithm? Professor Pangloss explains: 'Since we have been fooled by those guys from mainstream academia, we will not show how our procedure works! It is even impossible to rescue its inner working, because this was written on a sheet of ice that Professor Buraninev applied on his vodka research!'

But... what about computer codes and simulations? Again, Professor Pangloss has an explanation: 'Our codes, simulations, and computers were all eaten out by a blue whale! This animal is very common by this side of the world and even ate my favourite grad student!'

It seems that evolutionary computation will need to wait a little more to be completely formalized and put on formal grounds!"

Professor Schwefel's Keynote At EvoStar

Our blog friend Juan Julián Merelo Guervós is posting about the EvoStar 2008 event.

His first post is a short summary concerning the opening keynote given by Professor Hans-Paul Schwefel, see it here.

His recent post summarizes a little bit more the keynote presented by Professor Schwefel, see here.

Very interesting posts!

Upon the future of evolutionary computation, Professor Schwefel said that "the unexpected should be expected".

Professor Schwefel also mentioned a situation when a paper reviewer commented the following: "Why should other optimization algorithms be necessary?" This comment is strongly correlated to traditional optimization methods and the development of the evolution strategies, since some researchers - along the 1960s and 1970s - thought that gradient-based methods and others techniques (such as Gauss-Seidel method) were all the stuffs they needed.

He also spoke of the application of evolution strategies (under the form of the so-called experimental optimization) to a two-phase flashing nozzle optimization, which was performed without computers! Even today, it is not possile to calculate what happens within such nozzle: thermodynamically far away from equilibrium, drag between slow water droplets and fast steam, three-dimensional turbulent boundary layer with liquid sublayer, supersonic behind nozzle throat, etc.

There were several challenges to be overtaken when preparing the grounds to the evolution strategy, such as self-adaptation, non-elitist selection method, parallelism, and so on. He obtained inspiration in natural systems to implement solutions to those problems.

An important comment that evolutionary computation researchers should pay attention:

"He's commented that evolutionary algorithms are getting less bio-inspired in time, this is not good or bad, but it's more interesting for him to look at models than to have super-tweaked ultra-tuned purportedly bio-inspired algorithms."