|
Author
|
Topic: Speciation
|
RBH
Member
Member # 380
|
posted 28. April 2004 16:25
That representations of "latent" information can be extracted from non-latent, overt data sets is a statement about implicit structure of the data (possibly adventitious pseudo-structure due perhaps to sampling artifacts), but is not necessarily a statement about the 'reality' of the implicit or latent variables. They provide hypotheses about that, but not confirmation of it. The perpetual bane of my working life is discriminating between derived representations that are artifactual and a product of adventitious sampling and should therefore be discarded, and what's "real" and can therefore be depended on in risking client money.
That one can construct representations of latent information, as in training the hidden layer(s) in a NN backprop procedure, means that those representations are accessible to environmental feedback and are thus preservable by a process operating on the 'outside' explicit information, as natural selection does. That in fact is how evolutionary theory conceives the relationship between the fact that phenotypes are what are directly exposed to selection while simultaneously holding that the heritable information-bearing entities are genetic representations that are not themselves directly exposed to selection.
My original point therefore stands: evolutionary theory provides a mechanism for conserving representations of actively used information, be it explicit or latent, while Intelligent Design of the front-loading variety requires that an information representation, again either latent or explicit, that is unused for billions of years must nevertheless be somehow protected from the vicissitudes of thermodynamic battering in the absence of any hint of a mechanism for providing that protection. Hidden information that is (more or less) continuously utilized, as in the hidden layers in a NN in training, is indirectly accessible to environmental feedback and therefore may be conserved through time.
With respect to "antiquated" sources, the question is whether their points are currently cogent. That has to be established if they are to be taken as addressing current questions, not merely assumed. That Mivart's criticism of Darwin was or was not valid in the 19th century is irrelevant to current questions unless one can show that current evolutionary theory is identical (in the criticized respect) to OoS and that Mivart's criticism was valid in the first place. And Darwin did address Mivart's critique in the 6th edition of OoS.
RBH
[ 28. April 2004, 16:29: Message edited by: RBH ]
IP: Logged
|
|
nosivad
Member
Member # 767
|
posted 28. April 2004 17:04
Pim, you are suggesting that I have rejected evolutionary theory. That is absolute nonsense. I have, with many others, rejected the Darwinian model based on random mutation coupled with omnipotent Natural Selection. It is utterly without foundation. The pathways which may have existed in the evolution of the eye, or more accurately eyes, have absolutely nothing to do with the mechanisms by which such transformations took place. If there was ever an example of the lack of intermediates it is the vertebrate eye. Every example you offer is compatible with an emergent, prescribed evolution in which chance played no necessary role whatsoever. Furthermore, the reverse speciation recently demonstrated in yeast completely supports the Prescribed Evolutionary Hypothesis. Scrambling the chromosome has produced a new genotype with absolutely no participation by either mutation or selection. I am now of the opinion that the environment has played a very minor role in evolution. It has been, past tense of course, an emergent phenomenon determined by law as Bateson, Grasse and Berg have each independently proposed. After eliminating both Darwinism and Lamarckism there remains no other explanation.
IP: Logged
|
|
Pim van Meurs
Member
Member # 541
|
posted 28. April 2004 17:17
Nosivad: Pim, you are suggesting that I have rejected evolutionary theory. That is absolute nonsense. I have, with many others, rejected the Darwinian model based on random mutation coupled with omnipotent Natural Selection. It is utterly without foundation.
If Nosivad wants to argue the strawman that natural selection is omnipotent or even the only evolutionary mechanism then he is presenting another strawman. You have rejected without much foundation in my opinion, Darwinian theory.
Nosivad: After eliminating both Darwinism and Lamarckism there remains no other explanation.
That of course is based on two problems, Darwinism has not really been eliminated and it proposes a fallacy that there remain only three viable candidates.
I have shown beautiful examples of transitionals for the evolution of the ear and shown how darwinian pathways can explain the evolution of the eye. Some good explanatory websites exist that explain the evolution of the eye
PBS
Lindsay
Uncovering The Ancestry of A Complex Organ, The Eye
Much of Nosivad's rejection of Darwinian theory seems to be based upon the idea that natural selection is the only relevant mechanism and that variation through mutation is the only source of variation. A careful reading of Darwin would put to rest any such interepretation of his work as erroneous.
Other relevant research
Chromosomal evolution in Saccharomyces
quote:
Ian Roberts and Steve James, with colleagues from the Universities of Manchester and Oxford are trying to understand how chromosomes evolve and which are the critical changes in genetic material that result in the evolution of new species. They think several types of change to DNA are involved in the formation of new species; the most important was thought to be chromosomal rearrangements - where a region of DNA swaps between the arms of a pair of chromosomes. However, in this paper they show that this process alone was not responsible for producing new species of yeast.
and
Nature. 2000 May 25;405(6785):451-4. Chromosomal evolution in Saccharomyces.
Fischer G, James SA, Roberts IN, Oliver SG, Louis EJ.
quote:
The chromosomal speciation model invokes chromosomal rearrangements as the primary cause of reproductive isolation. In a heterozygous carrier, chromosomes bearing reciprocal translocations mis-segregate at meiosis, resulting in reduced fertility or complete sterility. Thus, chromosomal rearrangements act as a post-zygotic isolating mechanism. Reproductive isolation in yeast is due to post-zygotic barriers, as many species mate successfully but the hybrids are sterile. Reciprocal translocations are thought to be the main form of large-scale rearrangement since the hypothesized duplication of the whole yeast genome 10(8) years ago. To test the chromosomal speciation model in yeast, we have characterized chromosomal translocations among the genomes of six closely related species in the Saccharomyces 'sensu stricto' complex. Here we show that rearrangements have occurred between closely related species, whereas more distant ones have colinear genomes. Thus, chromosomal rearrangements are not a prerequisite for speciation in yeast and the rate of formation of translocations is not constant. These rearrangements appear to result from ectopic recombination between Ty elements or other repeated sequences
Nosivad: Every example you offer is compatible with an emergent, prescribed evolution in which chance played no necessary role whatsoever.
Without really any details as to what is meant by prescibed evolution, it is hard to argue that there is anything not compatible with it. The notion of teleology is merely an addition which of course is compatible with evolutionary theory which by virtue of its mechanisms is teleological in nature.
Ayala defines it as follows
"Explanation by design, or teleology, is "the use of design, purpose, or utility as an explanation of any natural phenomenon" (Webster's Third New International Dictionary, 1966)"
Ayala
quote:
Similarly features of organisms are teleological as well: a bird's wings are for flying, eyes are for seeing, kidneys are constituted for regulating the composition of the blood. The features of organisms that may be said to be teleological are those that can be identified as adaptations, whether they are structures like a wing or a hand, or organs like a kidney, or behaviors like the courtship displays of a peacock. Adaptations are features of organisms that have come about by natural selection because they serve certain functions and thus increase the reproductive success of their carriers.
[ 28. April 2004, 17:52: Message edited by: Pim van Meurs ]
IP: Logged
|
|
warren_bergerson
Member
Member # 262
|
posted 28. April 2004 17:40
Quote RBH: That in fact is how evolutionary theory conceives the relationship between the fact that phenotypes are what are directly exposed to selection while simultaneously holding that the heritable information-bearing entities are genetic representations that are not themselves directly exposed to selection.
It will be noted this all happens with many to many, dynamic phenotype to genotype mappings which can not be explicitly defined or demonstrated.
IP: Logged
|
|
Pim van Meurs
Member
Member # 541
|
posted 28. April 2004 18:02
YOu may want to check out the work by Toussaint in this area.
quote:
However, in natural evolution mutation operators are not designed by some intelligence. A central question arises: What does it mean to “learn” about the problem structure and exploit it? How in principle can evolution realize this? The answer we will give is that the implicit process of the evolution of genetic representations allows for the self-adaptation of the “search strategy” (i.e., the phenotypic variability induced by mutation and recombination). To some degree, this process has been overlooked in the context of evolutionary algorithms because complex, non-trivial (to be rigorously defined later) genetic representations (genotype phenotype mappings) have been neglected by theoreticians. This chapter tries to fill this gap and propose a theoretical framework for evolution in the case of complex genotype-phenotype mappings focusing at the evolution of phenotypic variability. The next section lays the first cornerstone by clarifying what it means to learn about a problem structure.
or On the evolution of phenotypic exploration distributions
Evolvability and robustness are both linked through neutrality
Also The Use of Neutral Genotype-Phenotype Mappings for Improved Evolutionary Search R Shipman and M Shackleton Volume 18, Issue 4 of BT Technology Journal
quote:
In natural systems, the organism or phenotype is the result of a complex developmental process that is played out as the genetic information is interpreted. This is in stark contrast to many artificial evolutionary systems in which the phenotype is represented directly in the genetic information and there is no such development. As well as overcoming the obvious practical impossibility of directly specifying an organism in the genotype, the developmental process may yield other desirable properties. One such property is neutrality in which many genotypes develop into the same phenotype. This paper examines the effect of neutral genotype-phenotype mappings on artificial evolutionary systems through examination of an abstract redundant mapping based on a random Boolean network (RBN). It then goes on to examine the genotype-phenotype mapping within a planning tool that evolves instructions for growing telecommunications networks. It is demonstrated how the right kind of redundancy has the potential of significantly aiding the evolvability of a system.
In this interesting thesis work, I found the following statement
quote:
More recent studies of evolution take into account some aspects of biological evolution
that are typically lacking in traditional studies, e.g. the multi-level character of biological systems (Hogeweg, 1994; Maynard Smith & Szathm´ary, 1995a,b; Hogeweg, 1998), self-structuring of biological systems and its consequences for evolving systems (Hogeweg & Hesper, 1991; Boerlijst & Hogeweg, 1991; Savill et al., 1997), non-linear genotype-phenotype mappings (Schuster, 1989; Huynen et al., 1993; Kauffman, 1993; Huynen & Hogeweg, 1994), or the occurrence of neutrality in genotype spaces (Huynen et al., 1996; Huynen, 1996; Van Nimwegen et al., 1999). These studies show that evolution is better viewed as a multi-level informatic process, i.e. a process in which “information
and novelty are created and processed across and between the self-structured multiple levels” (Savill, 1997).
I'd argue that most of the objections raised by Dembski against evolutionary algorithms are against those algorithms which fail to take into consideration the differences between natural evolutionary processes and their models. In other words once multiple (nonlinear) mappings and coevolutionary relationships are introduced, natural evolutionary processes appear to be quite different from the 'No free lunch' counterparts.
[ 28. April 2004, 18:22: Message edited by: Pim van Meurs ]
IP: Logged
|
|
RBH
Member
Member # 380
|
posted 28. April 2004 20:05
warren_bergerson wrote quote:
It will be noted this all happens with many to many, dynamic phenotype to genotype mappings which can not be explicitly defined or demonstrated.
Rather than a flat assertion, could we see a reference or two in support of this assertion? For example, does the work in evolutionary computation on genotype-phenotype mappings not speak to this question? Or the work in evo-devo? Is the gene-to-phene mapping a complete black box, impenetrable and totally unknown? I don't think so. Are these folks wasting their time characterizing the geneotype to phenotype mapping underlying Alzheimer's disease? How about these folks, working on Disease susceptibility genes for autism? Are they out to lunch? And is this review full of hot air?
How about providing some references to the professional research literature supporting your claim rather than a mere flat unsupported assertion from a non-professional, warren. C'mon: Demonstrate some familiarity with the literature of the discipline about which you regularly make claims, hm?
RBH
IP: Logged
|
|
|
|
warren_bergerson
Member
Member # 262
|
posted 28. April 2004 20:54
Genetic algorithm programs demonstrate that complex non-random search routines using precise logical structures including one to one static phenotype to genotype mappings can achieve purposeful goals in a predictable manner. We then admit that the type of mappings assumed in our GA programs don’t appear to apply in real biological systems.
Evolutionary and adaptive processes exhibit the same goal seeking behavior as genetic algorithm programs, but, apparently without the genetic determinism or static defined genotype to phenotype mappings assumed in the GA’s. Some people might describe this as a logical inconsistency between evolutionary theory and evolutionary mathematics. Some people are obviously content to ignore the inconsistency.
IP: Logged
|
|
Pim van Meurs
Member
Member # 541
|
posted 28. April 2004 21:00
Warren: Some people might describe this as a logical inconsistency between evolutionary theory and evolutionary mathematics. Some people are obviously content to ignore the inconsistency.
Why this suggestion, which flies contrary to what RBH and others have argued here, that people are 'content to ignore the inconsistency'? In fact I argue that your 'inconsistency' is one of your own creation and does not accurately reflect the arguments put forward by RBH and others on this group. Your statement not only fails to do justice to their arguments but also smell of a thinly veiled ad hominem attack.
You made a claim and all you can do is argue that there is an inconsistency because evolutionary computing, contrary to natural evolution, sometimes makes assumptions that are inaccurate in describing natural evolution?
First of all evolutionary algorithms are not always meant to accurately mimick natural evolutionary processes, secondly science makes approximations all the time and over time refines its approximations to better match reality. In other words when science finds out that evolutionary algorithms which more closely mimick actual evolutionary processes work better that this is somehow an inconsistency? In fact when science finds out that the nature of these genetic networks can be explained by simple processes (gene duplication) leading to 'scale free' networks, what relevance do your arguments really have? That science sometimes uses evolutionary algorithms in a way irrelevant to natural evolution? So what? Why not focus on the relevant applications of evolutionary computing lest we run the risk of attacking strawmen?
But back to supporting your claim. Can we expect some references?
[ 28. April 2004, 21:09: Message edited by: Pim van Meurs ]
IP: Logged
|
|
warren_bergerson
Member
Member # 262
|
posted 28. April 2004 21:11
Just read through the discussion. The inconsistency between one to one mappings in GA’s, and dynamic many to many mappings in biological systems comes from Rex and RBH.
IP: Logged
|
|
RBH
Member
Member # 380
|
posted 28. April 2004 21:15
warren_bergerson wrote quote:
Evolutionary and adaptive processes exhibit the same goal seeking behavior as genetic algorithm programs, but, apparently without the genetic determinism or static defined genotype to phenotype mappings assumed in the GA's. Some people might describe this as a logical inconsistency between evolutionary theory and evolutionary mathematics. Some people are obviously content to ignore the inconsistency.
Actually, one would define it as a difference between the two, a difference that is known and is a topic of active research. I note that warren_bergerson ignored the research on genotype-phenotype mappings in both evolutionary computation and biology that was pointed to, and slid off to a different question. I ask again, are there references in the professional literature that can be adduced to support warren_bergerson's claims or are we stuck in a permanent armchair?
RBH
IP: Logged
|
|
Pim van Meurs
Member
Member # 541
|
posted 28. April 2004 21:19
I have read through the discussion and all I noticed was the statement that
"It will be noted this all happens with many to many, dynamic phenotype to genotype mappings which can not be explicitly defined or demonstrated. "
Something which seems to be contradicted by the evidence. If you are ready to introduce some relevant evidence here to support your claim, fine.
The 'inconsistency' as I see it is one of your own creation and does not represent the arguments presented by RBH nor Rex.
But if you care to support your claim of inconsistency and since you seem to be suggesting that you were refering to Rex and RBH may I assume that your claim
"Some people are obviously content to ignore the inconsistency. "
was refering to RBH and Rex? If so, are you sure you are correctly representing their arguments? I have read quite a bit of RBH's contributions and your claim seems to go against my impressions of what RBH is arguing quite effectively.
If anything RBH seems to be quite well versed in evolutionary algorithms and how they apply to natural evolution as well as their limitations. RBH's latest response to Warren further supports my interpretation.
Some examples
quote:
as to whether GAs, or more generally, evolutionary algorithms will develop to the point that they are helpful in understanding the mechanisnms of evolution, it's pretty clear that the answer is yes, since some of that kind of work has already been published. See here for some examples.
Link
quote:
lding a real operating evolutionary algorithm that has to survive in a complicated and dynamic real-world environment (we pipe variables from the outside world into our GAs in real time) is a whole different ball game from building toys like Dawkins' weasel program. Even Schneider's ev program is a pretty simple program, not a full-blown ecosystem, yet it demonstrates an important deficiency in Dembski's claims. A GA that has to sense, monitor, adapt to, and control processes in a real-world dynamic environment is no toy. Toys are fine for illustration, but they don't display the full power of a technique.
Link
or even RBH's comments on ISCID
[ 28. April 2004, 21:28: Message edited by: Pim van Meurs ]
IP: Logged
|
|
RBH
Member
Member # 380
|
posted 28. April 2004 22:04
To provide an illustration of the fact that genotype-phenotype mappings are actually attended to, this from a neat paper with an artistic bent. quote:
3.3 SUMMARY : GENOTYPE TO PHENOTYPE MAPPING
The genotype to phenotype mapping is capable of creating complex complete graphs of coupled components. The mapping has some key properties.
- Epistasis: The expression of specific elements of the genotype depends upon epistatic interactions with other genes. This enables the retention of complex modules as dormant genetic material in the population, which can become re-expressed later and potentially combined with other modules. This characteristic is found in the homeobox genes of real organisms, enabling the reuse of genetic complexes for eye and limb formation, common genes which are shared between flies, human beings and mice. The development of such complexes is considered to have been key to the explosion of diversity in the Cambrian era. The negotiation of dependency between expressed elements provides a basis for epistatic interactions between genes.
- Redundancy: Modules which are key can be re-expressed many times in the genotype of a specific organism, since the position in the genotype is not important. This ensures that these key components are not lost owing to mutation.
- Neutrality: Many changes in a genotype make no change in the phenotype, either because they remain unexpressed, (through epistasis) or because they do not change the couplings between components, (for example when a coordinate changes by a small amount, and the associations formed in the developmental space remain identical). This encourages diversity of dormant genetic material within the population, and supports the development of new modules through the crossover of distinct material between two members of the same species.
Now I'm done with warrenbergerson's armchair nonsense. Life is too short.
RBH
[ 28. April 2004, 22:08: Message edited by: RBH ]
IP: Logged
|
|
nosivad
Member
Member # 767
|
posted 28. April 2004 22:43
One might get the impression from recent posts that neoDarwinism is alive and well and explains everything about an event which has never been directly observed. That is nothing but wishful thinking. It has no explanatory power whatsoever because it fundamentally denies purpose and design, both of which are evident to the unbiased observer. Evolution, like the development of the individual from a single cell has been an entirely predetermined series of events, resulting from the ordered sequential derepression of an originally enormous stockpile of potentialities. There is now and never was any role for chance in determining those sequences. If there had been, evolution could never have occurred. Darwinism is the last and indeed the only refuge of the confirmed atheist.
IP: Logged
|
|
Rex Kerr
Member
Member # 632
|
posted 29. April 2004 01:08
quote:
The inconsistency between one to one mappings in GA’s, and dynamic many to many mappings in biological systems comes from Rex and RBH.
This is a simplification, not an inconsistency. Anyone who wishes, and invests the time in learning about the subject, can calculate bounds on the error induced by using the simplification, and come up with rules such as: this simplification breaks down if the coupling between phenotype and genotype is less that such-and-so.
The distinction between simplification and inconsistency is extremely important to science and engineering, and should not be mixed up.
quote:
It [neoDarwinism] has no explanatory power whatsoever because it fundamentally denies purpose and design, both of which are evident to the unbiased observer.
Function is evident to the unbiased observer. How does one distinguish between function (which is a description of what a process does) from purpose (which requires intent) and design (which in this context presumably requires a conscious designer)?
[ 29. April 2004, 01:10: Message edited by: Rex Kerr ]
IP: Logged
|
|
|
Topic Closed
Printer-friendly view of this topic
