Uncommon Descent Serving The Intelligent Design Community

Who thinks Introduction to Evolutionary Informatics should be on your summer reading list?

Share
Facebook
Twitter
LinkedIn
Flipboard
Print
Email

Return to product information Robert Marks sends these endorsements for Evolutionary Informatics:

(Note: It is surprisingly easy to read.)

···············
“An honest attempt to discuss what few people seem to realize is an important problem. Thought provoking!”

Gregory Chaitin, Ph.D.
Professor, Federal University of Rio de Janeiro
Eponyms: Kolmogorov-Chaitin-Solomonov Information Theory
Chaitin’s Number
Chaitin’s algorithm
Author of:The Unknowable
Meta Math!: The Quest for Omega
The Limits of Mathematics
Thinking about Gödel and Turing: Essays on Complexity
Algorithmic Information Theory.

···············

“Darwinian pretensions notwithstanding, Marks, Dembski, and Ewert demonstrate rigorously and humorously that no unintelligent process can account for the wonders of life.”

Michael J. Behe, Ph.D.
Professor of Biological Sciences , Lehigh University
Author of: Darwin’s Black Box
The Edge of Evolution

···············

“This is a fine summary of an extremely interesting body of work. It is clear, well-organized, and mathematically sophisticated without being tedious (so many books of this sort have it the other way around). It should be read with profit by biologists, computer scientists, and philosophers.”

David Berlinski, Ph.D.
Author of: The Devil’s Delusion, The Deniable Darwin and Other Essays, The King of Infinite Space: Euclid and His Elements

···············

“For decades and decades, the ubiquitous cultural lie is that Intelligent Design advocates do nothing but rehash old criticisms of evolutionary theory. They never present fresh, positive research that supports ID theory. Now repeating serious criticisms of evolution is very important, especially since the universities, state school boards, and the ACLU have guaranteed that students must never hear of the problems with evolutionary theory. Still, the ID movement must present positive research for its views, and since this has been done for years through a number of publications, it is now a sign of ignorance, intellectual bigotry and bad faith for people to perpetuate this cultural lie. It is itself a lie. But with the publication of the ground-breaking book, Introduction to Evolutionary Informatics, there is now a cutting-edge positive ID research volume that does fresh, heretofore unpublished (and un-thought of!!) ideas that get to the very deepest bottom of recent science that is not only relevant to the ID/Evolution debate, but actually devastates evolutionary theory at the ground floor. In my view, no one reading this book can continue to adopt Theistic Evolution on philosophical and scientific grounds alone. This is must reading for all believers and unbelievers interested in the debate, and Christians who are scientists have, I believe, a moral and spiritual duty to read this book. Though somewhat difficult, Marks, Dembski and Ewert have done a masterful job of making the book accessible to the engaged and thoughtful layperson. I could not endorse this book more highly.”

J.P. Moreland, Ph.D.
Distinguished Professor of Philosophy, Biola University,
Author of: The Soul: How We Know It’s Real and Why It Matters

···············

“With penetrating brilliance, and with a masterful exercise of pedagogy and wit, the authors take on Chaitin’s challenge, that Darwin’s theory should be subjectable to a mathematical assessment and either pass or fail. Surveying over seven decades of development in algorithmics and information theory, they make a compelling case that it fails.”

Bijan Nemati, Ph.D.
Jet Propulsion Laboratory
California Institute of Technology

···············

“Dr. Marks has been at the forefront of research on evolutionary algorithms for three decades. However, in 2007 his university removed the website of his Evolutionary Informatics group because his research was a threat to the status quo in evolutionary biology. Nonetheless, Dr. Marks and his colleagues continued to pursue research into the informational requirements of evolutionary algorithms, the result of which is found in this volume. If you want to know what information theory says about evolution, this is the volume to read.”

Jonathan Bartlett, Director
The Blyth Institute
Author Programing from the Ground Up
Building Scalable Web Applications Using the Cloud
Coeditor Engineering and the Ultimate: An Interdisciplinary Investigation of Order and Design in Nature and Craft
Naturalism and Its Alternatives in Scientific Methodologies

···············

“Introduction to Evolutionary Informatics is a lucid, entertaining, even witty discussion of important themes in evolutionary computation, relating them to information theory. It’s far more than that, however. It is an assessment of how things might have come to be the way they are, applying an appropriate scientific skepticism to the hypothesis that random processes can explain many observed phenomena. Thus the book is appropriate for the expert and non-expert alike.”

Donald Wunsch, Ph.D.
Mary K. Finley Missouri Distinguished Professor
Director of the Applied Computational Intelligence Lab
Missouri University of Science & Technology
IEEE Fellow, INNS Fellow
Past President of the International Neural Networks Society
Coauthor of Neural Networks and Micromechanics
Unified Computational Intelligence for Complex Systems Clustering

···············

“Evolution requires the origin of new information. In this book, information experts Bob Marks, Bill Dembski, and Winston Ewert provide a comprehensive introduction to the models underlying evolution and the science of design. The authors demonstrate clearly that all evolutionary models rely implicitly on information that comes from intelligent design, and that unguided evolution cannot deliver what its promoters advertise. Though mathematically rigorous, the book is written primarily for non-mathematicians. I recommend it highly.”

Jonathan Wells, Ph.D. Ph.D.
Senior Fellow, Discovery Institute
Author of: Zombie Science,
Icons of Evolution
The Myth of Junk DNA

···············

“When biologists finally come to terms with the fact that Darwinism was a long experiment in collective self-deception, the work described in this book will deserve much of the credit for putting things right.”

Douglas Axe, Ph.D.
Director of Biologic Institute
Author of Undeniable: How Biology Confirms Our Intuition That Life Is.
Coauthor of Science and Human Origins

···············

“Introduction to Evolutionary Informatics helps the non-expert reader grapple with a fundamental problem in science today: We cannot model information in the same way as we model matter and energy because there is no relationship between the metrics. As a result, much effort goes into attempting to explain information (and intelligence) away. The authors show, using clear and simple illustrations, why that approach not only does not work but cannot work. It impedes understanding of our universe. The picture that emerges from their work is of a universe that is at the same time more mysterious than we had been led to expect and more familiar.”

Denyse O’Leary, Science Writer.
Author/Coauthor of:
The Spiritual Brain: A Neuroscientist’s Case for the Existence of the Soul
By Design Or By Chance?: The Growing Controversy On The Origins Of Life In The Universe

···············
“Marks, Dembski, and Ewert have written a book summarizing in a very accesible way all of their research at the Evolutionary Informatics Lab for the last decade. If the blind watchmaker says “me thinks it is like a weasel”, they say “perhaps, but in order to see it you need these active-information glasses.” When the watchmaker is able to see with the glasses (and he needs them to be certain it is a weasel), he is not blind anymore. He is, like the programmer of an evolutionary algorithm, an intelligent designer with a very clear sight of his target. —‘Oh, yes, it was a weasel!’ “

Daniel Andrés Díaz Pachón, Ph.D.
Research Assistant Professor, Biostatistics, University of Miami

···············

“This is an important and much needed step forward in making powerful concepts available at an accessible level.”

Ide Trotter, Ph.D.
Trotter Capital Management Inc.
Founder:Trotter Prize & Endowed Lecture Series on Information, Complexity and Inference (Texas A&M)

···············

“Steampunk fiction anachronistically fuses Victorian steam powered technology into the digital age. Darwinism is ‘steampunk science.’ It is an analog-based Victorian relic trying to make its way in the digital information age. Darwin had no conception of the information problem facing any account of naturalistic evolution. Darwin’s 21st century successors certainly know about the problem, but as Marks, Dembski and Ewert demonstrate in Introduction to Evolutionary Informatics, in 2017 they are no closer to solving the problem than Darwin was in 1859. This lay-accessible introduction to the information issue and how it remains unsolved is absolutely essential to anyone who wants to understand how all life is fundamentally information-based, and how naturalistic evolutionary science has not come remotely close to solving the problem of how meaningful information can arise in the absence of intelligence.”

Barry Arrington, D.Jur.
Colorado House of Representatives (1997-1998)
Editor-in-Chief, UncommonDescent.com

···············

“One of the things Intelligent Design theorists do is take what is obvious to the layman, that unintelligent forces cannot do intelligent things, and state it in more rigorous, scientific terms, so that highly educated people can understand also. This book makes important contributions to that effort, using results and terminology from information theory.”

Granville Sewell, Ph.D.
Professor of Mathematics, University of Texas, El Paso
Author of: Computational Methods of Linear Algebra
In the Beginning: And Other Essays on Intelligent Design
Christianity for Doubters

···············

“A very helpful book on this important issue of information, which evolution cannot explain. Information is the jewel of all science and engineering which is assumed but barely recognised in working systems. In this book Marks, Dembski and Ewert show the major principles in understanding what information is and show that it is always associated with design.”

Andy C. McIntosh DSc, FIMA, C.Math, FEI, C.Eng, FInstP, MIGEM, FRAeS.
Visiting Professor of Thermodynamics, School of Chemical and Process Engineering, University of Leeds, LEEDS, UK. Adjunct Professor, Department of Agricultural and Biological Engineering. Mississippi State University, Starkville, Mississippi, USA

···············

People who don’t like the book still won’t.

But see also: Information theory is bad news for Darwin: Evolutionary informatics takes off

Comments
I started a thread at "The Skeptical Zone": Introduction to Evolutionary InformaticsDiEb
July 13, 2017
July
07
Jul
13
13
2017
02:16 PM
2
02
16
PM
PDT
PS: I guess such a μ_s would have to exist trivially, just normalize the P_s function.daveS
July 9, 2017
July
07
Jul
9
09
2017
09:13 AM
9
09
13
AM
PDT
DiEb, Thanks for the reference and the further explanation. If I understand #68 correctly, this shows that for some searches, there is no suitable distribution μ_s that completely characterizes the function P_s which gives the probabilities of locating the targets. Will there always exist a distribution μ_s and a constant of proportionality k such that P_s(T) = k*μ_s(T)? It seems that k = 10 might work in the example from #68, but I don't know if that holds in general.daveS
July 9, 2017
July
07
Jul
9
09
2017
06:23 AM
6
06
23
AM
PDT
PS: That lot of background work required to run on a machine includes the biological case that requires metabolic, self replicating machines capable of reading and acting on strings of genetic information. Just ponder protein synthesis if that is too vague or generic for you. And for representations try a von Neumann Kinematic Self Replicator with an integrated fabrication facility.kairosfocus
July 9, 2017
July
07
Jul
9
09
2017
05:33 AM
5
05
33
AM
PDT
F/N: On search in an "evolutionary" context, this from Talib S. Hussain may provide convenient context, as there have been attempts to cloud terms like search:
Researchers in many fields are faced with computational problems in which a great number of solutions are possible and finding an optimal or even a sufficiently good one is difficult. A variety of search techniques have been developed for exploring such problem spaces, and a promising approach has been the use of algorithms based upon the principles of natural evolution . . . . In a search algorithm, a number of possible solutions to a problem are available and the task is to find the best solution possible in a fixed amount of time. For a search space with only a small number of possible solutions, all the solutions can be examined in a reasonable amount of time and the optimal one found. This exhaustive search, however, quickly becomes impractical as the search space grows in size. Traditional search algorithms randomly sample (e.g., random walk) or heuristically sample (e.g., gradient descent) the search space one solution at a time in the hopes of finding the optimal solution. The key aspect distinguishing an evolutionary search algorithm from such traditional algorithms is that it is population-based. Through the adaptation of successive generations of a large number of individuals, an evolutionary algorithm performs an efficient directed search.
Of course, the dominant problem is that of the population of possible solutions to functional challenges, the vast majority are utterly non-functional and functional cases come in deeply isolated clusters in the space of possibilities. That is, the inherent architecture of the problem is directly analogous to the task of a discoverer with very limited resources seeking out islands in a vast ocean, without a map or guides to the islands. As a result, any approach that turns on blindly picking cases from the space and then hoping to cross breed good or relatively good performers then do a sample of a new generation of cases to repeat runs into the problem of almost certainly fruitlessly exhausting search resources without ever landing on a shoreline of function. Where, Islands of function exist as the right components, correctly oriented and arranged then coupled -- this can be represented in bit strings in some description language -- must be in place with only limited tolerance, for function to emerge. On having a reasonably efficient description language [say, AutoCAD as a yardstick], the set of possibilities for 500 - 1,000 bits is 3.27*10^500 to 1,07*10^301. The first vastly exceeds search capability of the sol system, the second, that of the observed cosmos. So, we instantly see that evolutionary algorithms critically depend for success on careful and information-rich fine tuning. Fine tuning that puts the initial population in a favourable context where some function at least emerges so that improved cases can be rewarded by cross-breeding and then contributing to the next generation of cases, leading to a hoped for pattern of improved performance. Which of course has to be measured or observed in the computer but is analogous to survival and reproductive success of the fittest; never mind issues on circularity that often obtain. This involves what Dembski et al term active information. But in fact there is a lot more of such work going on in the background to get the schemes to work on a given machine. In short, the FSCO/I problem is real and is critical, regardless of real or imaginary flaws in any particular analysis. Unless this is acknowledged to be a key challenge, the debates we see pivot on begging the dominant question. KFkairosfocus
July 9, 2017
July
07
Jul
9
09
2017
04:38 AM
4
04
38
AM
PDT
DiEb, you have repeatedly tried to establish a perception of irrelevance by dismissive remarks. I have responded on substance showing that just the opposite is the case, kindly note 41 ff above especially and the remarks I made overnight when you went back to the irrelevancy claims yet again. Also, note the key remark cited at 63 above, do you have a reasonable, empirically warranted response that shows otherwise? Never mind whatever flaws you may find or think you find, is this substantial point groundless, on what basis of evidence that we observe -- key word -- blind search to create copious functional information and systems that can profitably employ it? Failing at this point, critiques will boil down to straining out gnats while swallowing camels. In short, the issue is due balance that reckons with the main issue, rather than trying to nip at heels and creating a false impression that this is of no consequence, there is no real problem here. The problem of origin of FSCO/I by blind search is a serious challenge, one that is at the heart of OOL and origin of body plan claims through the evolutionary materialistic school of thought. The evidence in hand is, the only actually empirically warranted cause of such FSCO/I is design, to the point that it is a good sign of design as cause. That is what needs to be addressed squarely. But for years, I find that objectors will do almost anything but this. KFkairosfocus
July 9, 2017
July
07
Jul
9
09
2017
03:29 AM
3
03
29
AM
PDT
@KF: "Per fair comment, my remarks have been shown to be relevant to the issues at stake, to the wider corpus of work by Dembski et al, and to the specific content of the book under discussion as descried by one of its authors." Fair comment? What are you talking about? I'm really lost...DiEb
July 9, 2017
July
07
Jul
9
09
2017
03:20 AM
3
03
20
AM
PDT
@daveS: If you have to guess a number between 1 and 100, and you may guess 10 times, then - in the language of Wolpert and Macready - this is akin to optimizing the characteristic function of the number. The set of characteristic functions of single elements of the set {1,...,100} is closed under permutation, so the FLT can be applied. The result: In this case, no strategy is on average over all targets better than 10 random guesses (without repetition) - in 10% of the searches you are successful. Now, DEM say that a search " induces a probability distribution μ_s on Ω that entirely characterizes the probability of S successfully locating the target T" If I take the average about over the targets, I get .01 * Σ_{T \in Ω} μ(T) = .01 = 1% That is surprising! BTW: the definition of a search in their earlier paper "The Search for a Search: Measuring the Information Cost of Higher Level Search" leads to the 10% figure, too - it just has other problems....DiEb
July 9, 2017
July
07
Jul
9
09
2017
03:09 AM
3
03
09
AM
PDT
F/N: On search, here are my remarks above in reply to questions, which appear to have been ignored in the repetitive demand for definition of search -- something that seems strange to me given the context of Darwinian evolution and its long history of computer simulation using searches across a space filtered by some sort of quality function, typically a "fitness function":
17 kairosfocusJuly 5, 2017 at 9:00 am DiEb, search is in effect a synonym for sampling, from flat random to extremely biased. the point is, search and space and in many cases initial point — think random walk — have to be well matched for success such as high function based on configuration to emerge. Search is thus inherently a complex topic, and to pretend that oh it can be made simplistic, is to fail to address reality. Further, search on material resources, i.e. search with resource and temporal costs [and monetary ones often] is constrained by that reality, so that scope of space to be searched is a highly relevant issue. I am sure you have seen decision analysis on whether anticipated cost of further investigation is likely to be worthwhile on reward likely to be obtained. This is a part of the bounded rationality problem in decision-making. One of the constraints here is likely scope of space of possibilities and available resources to search, esp if the search is by blind chance and/or mechanical necessity. A space of 4 bits is readily exhaustively searched, one of 500 bits is infeasible on the gamut of the sol system, and 1,000 on the gamut of the observed cosmos. And this is directly relevant to search for viable configs for origin of cell based life in a Darwin warm pond or the like. Boiled down, as a search is a sample, a search for a golden search comes from the power set of the original space, which is exponentially harder. So, direct first level search is the practical upper limit. On this, when we deal with complex configuration based function, blind search is maximally unlikely to access relevant function, but we know — as your comment illustrates — that design routinely and rapidly generates solutions. So, we can readlily see how a complexity threshold of 500 – 1,000 bits becomes a good test for design as most credible causal explanation. Indeed, of trillions of observed cases of such FSCO/I, never has there been an observation of such by blind chance and mechanical necessity. We routinely produce results by design that exceed the limit. So, no this is not something to be lightly dismissed. KF 21 kairosfocusJuly 5, 2017 at 10:06 am Kindly, explain what is wrong or irrelevant with: “search is in effect a synonym for sampling, from flat random to extremely biased. the point is, search and space and in many cases initial point — think random walk — have to be well matched for success such as high function based on configuration to emerge. Search is thus inherently a complex topic . . . ” KF PS: Search: Se_x = Sa_x (Space_y) | start place_z –> Success filter. Where, a particular search and a particular space must be given, also initial point, which can be forced or itself a prior random or intentional choice. [and more . . . ]
Per fair comment, my remarks have been shown to be relevant to the issues at stake, to the wider corpus of work by Dembski et al, and to the specific content of the book under discussion as descried by one of its authors. That SHOULD be enough for a reasonable discussion. KF PS: Perhaps terms like: Configuration space: https://en.wikipedia.org/wiki/Configuration_space State Space: https://en.wikipedia.org/wiki/State_space_(physics) or Phase Space: https://en.wikipedia.org/wiki/Phase_space are unfamiliar, so these should be helpful first level links, noting of course the limitations of that site. For instance, state space is commonly used for control systems to denote an abstract space that defines state. Configuration space is in effect synonymous. Microstate and macrostate are terms in statistical thermodynamics. Gibbs and Boltzmann are key pioneers.kairosfocus
July 8, 2017
July
07
Jul
8
08
2017
05:48 PM
5
05
48
PM
PDT
DiEb, with all due respect, I think you need to go look in a mirror, starting with the manner of your reply at 16 above and the pattern you have sustained since. Especially, in the aftermath of my highlighted and annotated citation from Marks at 41 above. KF PS: On the latest tangent on probability distributions, I think you will see a reasonable summary of why I spoke as I did here, http://www.itl.nist.gov/div898/handbook/eda/section3/eda361.htm Particularly note the generic sense involved, antecedent to particular models and functions.kairosfocus
July 8, 2017
July
07
Jul
8
08
2017
04:12 PM
4
04
12
PM
PDT
PPS: Cf that to my first comment above, which seemed to set you off on a campaign starting with, oh, give a precising definition of search:
15 kairosfocusJuly 5, 2017 at 6:58 am DiEb, the problem is it is not hard for a search space to become so large that reasonable search becomes impossible. Under those conditions, match of strategy, start-point, and specifics of space become important, especially if a space does not have the sort of convenient pointing slopes that some use to convey a misleading impression of the likelihood of searches. Where, 500 bits specifies 3.27*10^150 possibilities and 1000, 1.07*10^301. The first exhausts sol system resources and the latter, those of the observed cosmos. And of course, with a suitable description language, all searches come down to searches on binary spaces of suitable bit depth. KF
kairosfocus
July 8, 2017
July
07
Jul
8
08
2017
04:09 PM
4
04
09
PM
PDT
@KF: sorry, I was not aware that I was the one who played tangential word games...DiEb
July 8, 2017
July
07
Jul
8
08
2017
04:09 PM
4
04
09
PM
PDT
PS: BTW, money shot from this new book -- and again a key issue that has been resisted for years by objectors in and around UD:
Design is an inherently iterative process requiring domain intelligence and expertise. Domain knowledge and experience can be applied to the procedure to decrease the time needed for a successful search. Because of the exponential explosion of possibilities (i.e. the curse of dimensionality), the time and resources required by blind search quickly become too large to apply. Undirected Darwinian evolution has neither the time nor computational resources to design anything of even moderate complexity. External knowledge is needed. Neither quantum computing nor Moore's law makes a significant dent in these requirements. [Marks, Dembski, Ewert, Introduction to Evolutionary Informatics, World Scientific, 2017, p.59.]
kairosfocus
July 8, 2017
July
07
Jul
8
08
2017
04:00 PM
4
04
00
PM
PDT
DiEb, if you want to play tangential word games Mathematics is exactly NOT a science. In fact I used words in a reasonable manner, in the specific context of the Gibbsean distribution of probabilities across microstates; which extends into information metrics. But then you said already that you were picking up snippets and commenting -- cf 41 and 46 above. And in turn, that was in context of what seems to have gone poof once I cited, highlighted and noted on Marks' remarks on the context of the current book. Which showed that my earlier remarks were in fact relevant. KFkairosfocus
July 8, 2017
July
07
Jul
8
08
2017
03:54 PM
3
03
54
PM
PDT
@daveS: "Does the distribution P_s(k) that DEM talk about then give you the probability that search S finds the target, given that the target was k?" Exactly. As DEM write in "A General Theory of Information Cost Incurred by Successful Search":
Applying the discriminator Δ to this random search matrix thus yields an Ω-valued random variable Δ(S), which we denote by X_S . As an Ω-valued random variable, X_S therefore induces a probability distribution μ_s on Ω that entirely characterizes the probability of S successfully locating the target T. In this way, an arbitrary search S can be represented as a single probability distribution or measure μ_s on the original search space Ω. This representation will be essential throughout the sequel.
DiEb
July 8, 2017
July
07
Jul
8
08
2017
03:46 PM
3
03
46
PM
PDT
@KF: Mathematics in an exact science where words have meanings. You cannot simulate mathematics with a torrent of unspecified words.DiEb
July 8, 2017
July
07
Jul
8
08
2017
03:34 PM
3
03
34
PM
PDT
DiEb, I am not using distribution in the sense of a closed form function we can readily state or statistically readily identify but in the simple descriptive sense that everything from flat random to some things certainly so and others certainly not so comprise distributions of probabilities. Notice, hitherto I have not spoken to probability distribution functions but to the sort of thing we may see with a fair vs a loaded die to use a simple case in point. The onward return to focal issues is as I already stated. KFkairosfocus
July 8, 2017
July
07
Jul
8
08
2017
03:27 PM
3
03
27
PM
PDT
DiEb, The distribution δ_7 is not at all what I had in mind for this search. I was thinking more in terms of a distribution which guides the selection of samples. Does the distribution P_s(k) that DEM talk about then give you the probability that search S finds the target, given that the target was k?daveS
July 8, 2017
July
07
Jul
8
08
2017
03:22 PM
3
03
22
PM
PDT
Thanks for the reply, DiEb.daveS
July 8, 2017
July
07
Jul
8
08
2017
02:47 PM
2
02
47
PM
PDT
@KF: there is a difference between "having probabilities for events" and the concept of a "probability distribution"! If I give you two chances to guess a number randomly chosen from the set {1,2,3}, you could guess each event correctly with probability 2/3 - that does not give you a "probability distribution" on {1,2,3}....DiEb
July 8, 2017
July
07
Jul
8
08
2017
02:44 PM
2
02
44
PM
PDT
daveS: "What precisely is the distribution in the example I cited? I’m not necessarily doubting what you are saying, but I want to be clear about what it is." The method leads to identifying the number "7" with probability 1. According to DEM, the induced probability distribution is $\delta_7$. What if 8 was the target? One would think that you should find it with probability 1, too. No, not according to DEM: you will find it with $\delta_7 (8) = 0$, ergo never. But wouldn't you be let to the number 8 by the answers of the oracle/values of the function? No, as DEM separate the minimum of the function and the actual target: you have no way to identify the target within your search. That's why I bring up the complete search: Checking every single number should reveal the target. But not in the universe of DEM: That's what I call paradoxical. With DEM you have to produce your candidate for a target, and they will tell you whether it was the target afterwards. I'd define a search and a search algorithm along the lines of Wolpert and Macready in their paper on "No Free Lunch Theorems For Optimization": in most of the search problems you have a space and function on this page with a range of (at least partially) ordered values. The target is the element where the function reaches its optimum, so optimization and search are two sides of the same medal. For an unassisted search, the function is just the characteristic function of the target, other problems provide more complex functions: for the WEASEL, you have the Hamming-distance. You can think of the Traveling-Salesman-Problem (TSP) as a search for the shortest way through all cities. In the first two examples, you know the optimum of the function beforehand, so you can identify your target during the search. For the TSP, the optimum is not known upfront. But at least theoretically, you can enumerate all possible ways and identify the optimum, ergo find the target.DiEb
July 8, 2017
July
07
Jul
8
08
2017
02:37 PM
2
02
37
PM
PDT
DS, I am not interested in whether a distribution has a given closed form or can be reconstructed from stochastic studies, the relevant point is, once you select or sample from a set, some distribution will be there. Even, the crude, I will never sit in row 13 -- oops, 12A -- on a flight. (I think some people will refuse to fly rather than sit in that row.) Back to focal issues, sampling selects a subset (up to the point it's a census) and the set of all searches of a config space will be the set of its subsets. This gives us a simple way to see how search for a golden search will be exponentially harder than a direct search. Back on more core focus, when we have a case where blind search of a config space is forced to be extremely sparse (by way of resource exhaustion of sol system or observed cosmos) it then shows why deeply isolated islands of function will be effectively unobservable -- by overwhelming statistical weight of non-functional/meaningless states. This then leads to the dominance of fruitless needle in haystack blind search challenge over hoped-for hill climbing by incremental change of members of populations through recombinations, mutations etc. And, FSCO/I will naturally come in that needle in haystack pattern as components have to be pretty much right, have to be properly oriented, and have to be correctly arranged and coupled for coherent function to emerge. KFkairosfocus
July 8, 2017
July
07
Jul
8
08
2017
01:52 PM
1
01
52
PM
PDT
KF,
DS, any search or selection process on a set will impose a probability distribution. Certainty is also a probability level. KF
What precisely is the distribution in the example I cited? I'm not necessarily doubting what you are saying, but I want to be clear about what it is.daveS
July 8, 2017
July
07
Jul
8
08
2017
11:50 AM
11
11
50
AM
PDT
DS, any search or selection process on a set will impose a probability distribution. Certainty is also a probability level. KFkairosfocus
July 8, 2017
July
07
Jul
8
08
2017
11:37 AM
11
11
37
AM
PDT
DiEb & KF,
One of the most problematic sentences is on page 173: “We note, however, the choice of an [search] algorithm along with its parameters and initialization imposes a probability distribution over the search space”.
This is over my head, and I haven't read the book, but I'm trying to understand what probability distribution would arise in the case of a binary search. Here's an example lifted from the wikipedia page on binary search. The goal of the search is to find the "7" in the sorted array. I presume that the "search space" consists of the set of numbers in the original array. Does the algorithm naturally define a probability distribution on that space?daveS
July 8, 2017
July
07
Jul
8
08
2017
07:13 AM
7
07
13
AM
PDT
A separation of the search (binary strings) and solution spaces (sentences) which removes the necessity to exclusively adopt a variable-length genetic algorithm (or even any evolutionary algorithm!) as is standard in GE as the search engine. The search operators of the evolutionary algorithm themselves ... - Natural Computing Algorithms
DiEb demands that they define "search"! Or not.Mung
July 7, 2017
July
07
Jul
7
07
2017
09:45 PM
9
09
45
PM
PDT
@DiEb: I’m looking forward to your definition of the term “search”!Mung
July 7, 2017
July
07
Jul
7
07
2017
09:11 PM
9
09
11
PM
PDT
This leads us to the notion that the process of problem solving can be viewed as a search through a potentially huge set of possibilities to find the desired solution. Consequently, the problems that are to be solved this way can be seen as search problems. - Introduction to Evolutionary Computing
Maybe they need to define 'search.' If not, why not? DiEb, you sent them a demand letter, didn't you?Mung
July 7, 2017
July
07
Jul
7
07
2017
09:09 PM
9
09
09
PM
PDT
DiEb, I am looking forward to some pretty graphs from you! Maybe those pretty graphs you post over at TSZ don't depend on search though. Why don't you define search?Mung
July 7, 2017
July
07
Jul
7
07
2017
07:58 PM
7
07
58
PM
PDT
DiEb, The above directly shows the relevance of considerations above to the purpose and issues discussed in the book, AS WAS INDICATED BY AN AUTHOR. Second, sufficient description and symbolisation of search has been given long since in this thread. All I will add here, is that in a context of self-replicating entities with related metabolic and other functions [the von Neumann Kinematic self replicator with an integral functioning entity is a picture], success filtering would be by reproductive success across generations in relevant environments. And if you wish a relevant information structure, try DNA and changes by mutation etc. The net result will clearly be that even moderate quantities of information will not emerge by such a blind search process. Which, starting from some initial case -- that already is a huge jump start -- will be by a random walk process. KF PS: I particularly point you to 41, which is in fact mostly Marks, not me. If yoiu refuse to read more than a couple of hundred words by an author then fail to perceive the connexions between this book and the corpus of work over years, thence relevance of my own remarks, that failure is obviously of your own making. I suggest, start from 41, an approximation to a lost comment.kairosfocus
July 7, 2017
July
07
Jul
7
07
2017
05:32 PM
5
05
32
PM
PDT
1 2 3

Leave a Reply