Antievolution.org :: Antievolution.org Discussion Board The Critic's Resource on Antievolution

Antievolution.org Discussion Board > All About Antievolution > Intelligent Design > More stupidity from the mind of the master

Tom Ames Posts: 238 Joined: Dec. 2002 (Permalink) Posted: Feb. 10 2004,16:08    OK, this is just too much. Dembski is posting a blog (""Unfinished Thoughts") and for the life of me I can't tell whether he's being disingenuous or stupid. He writes: Quote Koza spoke on "biologically inspired computation." He is one of the key people in the field, and every few years edits a book whose title begins with Genetic Programming . . . (he's now up to Genetic Programming IV: Routine Human-Competitive Machine Intelligence, which appeared last summer).   In example after example, he described setting up a fitness/objective function and then seeking an optimal solution for it.  Interestingly, however, in all his examples the fitness/objective function always remained fixed. I therefore approached him after his talk and asked him whether he knew of any research in evolutionary computation that also changed/evolved the fitness/objective function in the search for an optimal solution to a computational problem. He replied no. I found this quite interesting since the Darwinists claim that one of the things that gives added power to Darwinian evolution is the fact that environmental fitness is dynamic rather than static, changing, for instance, in response to evolving organisms (this is supposed to be a key factor in evolving irreducibly complex biochemical machines). Yet such coevolving fitness landscapes, which I don't deny occur in biology, are absent from evolutionary computation. The Darwinist might want to interpret this difference thus: "Isn't it amazing that nature has given us a form of natural evolutionary computation which varies its fitness/objective functions and which therefore makes biological evolution that much more powerful than it is in silico? Just wait until computer scientists capture this feature of biological evolution. Just think of how much more powerful evolutionary computation will be then." My own view is rather different. The fact that fitness/objective functions that vary over time are not employed in biologically inspired computing, especially after all these years of genetic algorithms hype, tells me that they are not the key to solving interesting engineering problems. And if they can't do it in the engineering context, there's no reason to think they can do it in biological contexts. (emphasis added) Is he daft? "The Darwinist" might think no such thing. Try to transcend your parochial, sectarian worldview for a second, Bill, and think. What makes evolutionary computing work? It solves a problem. What happens when the fitness function is allowed to change during the course of the evolution? The problem changes. So if I am an engineer who wants to solve a problem should I use a procedure in which the problem is allowed to vary? Uh, no. The fact that engineers don't incorporate every phenomenon in evolutionary biology into their code says not a #### thing about the relevance of those feature to real world evolution. Seriously, is this guy capable of stringing two thoughts together? Or is he just so infatuated with the sound of his own shrill voice that he doesn''t bother looking at the meaning of what he says? Edited by Tom Ames on Feb. 10 2004,14:09 -------------- -Tom Ames

richard_wein Posts: 1 Joined: Feb. 2004 (Permalink) Posted: Feb. 21 2004,07:49    Good point, Tom. But in fact it's worse than this for Dembski. If an evolutionary algorithm features co-evolution (evolution of two or more competing entities) then the fitness function automatically changes as the evolution proceeds, because the fitness of an entity (its propensity for reproductive success) depends on the other entities it is competing against. And co-evolutionary algorithms certainly have been programmed (though I don't know if they've been used in engineering applications). Dembski even mentions one in his own book No Free Lunch, namely the checkers playing neural nets of Chellapilla and Fogel. So it is simply untrue that "fitness/objective functions that vary over time are not employed in biologically inspired computing", as Dembski claims. This just demonstrates, once again, Dembski's poor grasp of the concept of fitness functions. Indeed, this very error has already been pointed out at least twice, once by me and once by David Wolpert (co-discoverer of the NFL theorems), in our respective critiques of NFL. Here's what Wolpert wrote: "Perhaps the most glaring example of this is that neo-Darwinian evolution of ecosystems does not involve a set of genomes all searching the same, fixed fitness function, the situation considered by the NFL theorems. Rather it is a co-evolutionary process. Roughly speaking, as each genome changes from one generation to the next, it modifies the surfaces that the other genomes are searching." Wolpert refers here to "neo-Darwinian evolution of ecosystems", but clearly the same point applies to all co-evolutionary systems.

RBH Posts: 49 Joined: Sep. 2002 (Permalink) Posted: Feb. 21 2004,13:23    Some coevolutionary EAs out in the (near-) real world: A symbiotic evolutionary algorithm for the integration of process planning and job shop scheduling Abstract: Quote This paper addresses the integrated problem of process planning and scheduling in job shop flexible manufacturing systems. Due to production flexibility, it is possible to generate many feasible process plans for each job. The two functions of process planning and scheduling are tightly interwoven with each other. The optimality of scheduling depends on the result of process planning. The integration of process planning and scheduling is therefore important for an efficient utilization of manufacturing resources. In this paper, a new method using an artificial intelligent search technique, called symbiotic evolutionary algorithm, is presented to handle the two functions at the same time. For the performance improvement of the algorithm, it is important to enhance population diversity and search efficiency. We adopt the strategies of localized interactions, steady-state reproduction, and random symbiotic partner selection. Efficient genetic representations and operator schemes are also considered. While designing the schemes, we take into account the features specific to each of process planning and scheduling problems. The performance of the proposed algorithm is compared with those of a traditional hierarchical approach and an existing cooperative coevolutionary algorithm. The experimental results show that the proposed algorithm outperforms the compared algorithms. Multi-objective cooperative coevolution of artificial neural networks (multi-objective cooperative networks). Abstract: Quote In this paper we present a cooperative coevolutive model for the evolution of neural network topology and weights, called MOBNET. MOBNET evolves subcomponents that must be combined in order to form a network, instead of whole networks. The problem of assigning credit to the subcomponents is approached as a multi-objective optimization task. The subcomponents in a cooperative coevolutive model must fulfill different criteria to be useful, these criteria usually conflict with each other. The problem of evaluating the fitness on an individual based on many criteria that must be optimized together can be approached as a multi-criteria optimization problems, so the methods from multi-objective optimization offer the most natural way to solve the problem.In this work we show how using several objectives for every subcomponent and evaluating its fitness as a multi-objective optimization problem, the performance of the model is highly competitive. MOBNET is compared with several standard methods of classification and with other neural network models in solving four real-world problems, and it shows the best overall performance of all classification methods applied. It also produces smaller networks when compared to other models.The basic idea underlying MOBNET is extensible to a more general model of coevolutionary computation, as none of its features are exclusive of neural networks design. There are many applications of cooperative coevolution that could benefit from the multiobjective optimization approach proposed in this paper. and Applying Cooperative Coevolution  To Inventory Control Parameter Optimization.  Inventory control may not seem like a sexy application, but it's a large component of many firms' costs. RBH -------------- "There are only two ways we know of to make extremely complicated things, one is by engineering, and the other is evolution. And of the two, evolution will make the more complex." - Danny Hillis.

Track this topic	Email this topic	Print this topic

© 2002-2009 Antievolution.org Powered by Ikonboard 3.0.2a © 2001 Ikonboard