This should get some good discussion / denial!
652 thoughts on “Evolving Wind Turbine Blades”
Leave a Reply
You must be logged in to post a comment.
The Skeptical Zone
"I beseech you, in the bowels of Christ, think it possible that you may be mistaken."
It’s a waste of time. I’m already ignoring Frankie, should probably ignore Mung too.
Why? Do you have a point?
Who cares.
If GA’s are goal-driven, purposeful, designed, teleological, searches then they do not model or simulate evolution unless evolution is a goal-driven, purposeful, designed, teleological, search.
But, for whatever reason, people are dead set on insisting that GA’s are not what they are. Apparently without GA’s no one would believe in biological evolution. It’s an odd belief to have, but some people need to cling to it even though evolution did just fine for over a century before Holland.
The point was already made. You’re bright enough to figure it out. Now all you have to do is claim it’s wrong and say why.
Yes, I know. In your own unique way you agree with me while trying to pretend like you don’t. It’s ok.
Please do. The less nonsense I have to deal with here the better.
Not at all, Mung. The root problem seems to be .. you’re a bit thick.
Psst. Mung.
Psst. Richardthughes.
You’re bright, Richardthughes. You’ll figure it out.
I find the concept of “information” rather useless in these cases.
I would say that the system, rather a living thing, or a program, has tested the environment and learned from feedback.
The “learning” is not usefully depicted as information, because it usually isn’t stored as bits in a file or a memory chip. Instead, the learning is a change in the way the system behaves, and to borrow a useful concept from FMM, the change is usually integrated into the system. It is not necessarily in the form of data bits, but rather a change in the structure, with ramifications in multiple dimensions.
Fitness and reproductive success are the same thing. GA’s model that in the fitness function, which determines which elements reproduce and which don’t, AKA, natural selection. It’s not rocket science
Harking back to the absolute fitness thread: there’s a minimum level of absolute fitness. a living thing must be able to remain alive and to reproduce.
Fitness in a population is relative to others in the population. It could be due to relative absolute fitness (health), or it could be luck (avoiding asteroid impacts, or not being born in time of drought) or it could be having the right tail feathers to attract a mate.
Complicated, but not rocket science.
Mung:
You’re not, Mung. You probably won’t.
Yeah, I know. IRL it’s not that simple, specially with small populations I would think.
But in the context of GA’s, where compromises and simplifications are usually needed, I think it’s close enough. It would make no sense to simulate occasional accidents that eliminated some potentially beneficial mutations from the algo before they can reproduce
What are your thoughts on the matter? Are GA’s teleological and is evolution teleological? When you say “people” are you referring to anyone specifically?
You post many hypotheticals, if this, if that, should I respond? Why?
Post your GA code again, Mung.
And Richardthughes could design a GA that randomly assigns a fitness value to each genotype and then uses that value to determine the representation of each genotype in the next generation. We could call it a generic GA and then try to use it to solve problems.
What say we give it a shot and see how our generic GA does handling some actual problems?
Do you have a question?
Mung,
Clunk! Differential reproductive success causes a population to adapt (of course something causes differential reproductive success too…). The higher-reproducing genotype is, arguably, an adaptation. When it’s fixed (by selection), everyone has it. Everyone has adapted to the prevailing selective ‘wind’ (a metaphor).
Goodness, Mung, what would be the point of that? The whole point of “fitness” or reproductive success in biological evolution is that it is an environmental test that variants are subjected to. It is a non-random sorting.
Mung,
Because what you call it is absolutely vital. Vital, I say!
This is the most stupid thing so far. Fitness is a function of the genotype functioning in it’s environment. Making it random doesn’t evaluate anything.
Mung,
And yet the thing that inspired them doesn’t work! The selection isn’t natural; someone coded it!
Alan Fox,
His understanding is Gallienesque. No clue, but he’s arguing his corner!
Mung,
That would have been rather silly. He spent pages setting up a case of Artificial Selection.
“And now”, says he, ” I give you … er … Selection!”.
“Didn’t you just do that?”
“Yes, but this isn’t Artificial”.
“What is it then?”.
“Well … non-artificial I suppose”.
“What would you call that then?”.
” er … would you buy Natural?”
“OK. Honestly Chas, you aren’t making this easy. I think you need a decent editor”.
Totally obliterating the genotype/phenotype distinction and making a semantic mishmash out of adaptation. Arguably.
Arguably. Unless you’re talking about adaptation. Then not so much.
Mung,
If a weather forecasting program is used to predict the weather, it does not model or simulate the weather unless the weather is attempting to predict the weather.
Yea, well, don’t tell Alan Fox. Or Joe Felsenstein. Because God forbid that the selection in a GA be artificial [by the artifice of man] rather than natural.
Allan Miller,
Partly my fault. The distinctions between “natural”, artificial and sexual selection are, in my view, somewhat – well – artificial. The process, as far as phenotypes and genotypes are concerned is exactly the same. With the foresight of anticipating Creationist semantic wordplay, Mr D. could have called it environmental selection; but then plant and animal breeders and potential mates are aspects of the environment.
Designing an environment, like a wind tunnel or a flask of broth, is not designing an outcome, like an adaptation.
Mung,
Arguably not. You don’t think the process of adaptation involves change in genotype frequencies, nor that the phenotypic ‘adaptation’ is fundamentally built from those genotypes? Eeeenteresting.
No, the most fit are those that leave the most offspring. Full stop. Do follow along. The best reproducers reproduce best.
You’re beginning to understand. You left out an important point though. Fitness (reproductive success) is relative to the precise environment of the breeding population.
Alan Fox,
Yes, I’m inclined to agree. I just took the opportunity for a bit of levity, with the drink being upon me and all! (Elderflower cordial).
This is what you wrote upthread, remember.
The whole point of GAs is the non-random element of performance testing.
Exactly. But driving/ directing the population towards adapting to the environment, by actively searching the landscape and responding accordingly, is designing an outcome.
Environments change whereas the fitness function doesn’t (throughout the run). And in real life behavioral change trumps morphological change and doesn’t take long to implement.
And computers simulate weather and still they get the forecast wrong 🙂
Logic Fail. So many things you don’t understand: https://en.wikipedia.org/wiki/Post_hoc_ergo_propter_hoc
An impromptu encounter with a bottle of Glen Grant at New Year has led me to a resolution which is for the moment holding.
Exactly- performance towards a specified goal. All GAs do is emulate many people working on the same problem. Computers just do it faster and with fewer resources
Even those you said were in Lt. Data from Star Trek?
Then blame writers of fitness functions. You look to be stepping into map and territory issues.
LoL! Was that supposed to be a refutation? I was just stating a fact that seems to be overlooked. You look to be avoiding reality
Alan Fox,
Nice of you to avoid responding to the relevant parts of the post. I know why you did that
So we have a population of 1000
They perform some task, that is empirically measured.
Let’s call that Fitness.
The fitness is linearly rescaled – the lowest value now being 0 and the highest being 1.
We order each entity based on the new fitness, High to low.
Each entity has 2 chances at having an offspring if a random number is less than their current rescaled fitness.
Offspring may have a random mutation that affects their ability to perform the task
Once we have 1000 new critters, we begin again with those.
What do you think will happen, Mung?
I prefer ability to perform a task.
People disagree sometimes. You seem to see it as a sign of weakness when in fact it’s a sign of strength.
Does biology have a goal? If so, who is setting it? How do you know? How would you know if that guidance stopped?