Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Genetic Algorithm building a little car

wreck.devisland.net

Post reply on HN
Print view View original

41–50 of 59 posts

Re: Genetic Algorithm building a little car

#41

Post by kirse »

anthonyb wrote: post #38
kirse wrote: post #24

Earlier quoted context omitted.

How exactly is it evolutionary computing when there is a designer in the equation (programmer) who puts constraints on the algorithm by the simple act of coding it and running it on systems that he designed? This seems more like "computing to model the process of genetic adaptation". You can't label it evolutionary when there's a designer in the equation. The very definition of the evolutionary process does not allow…

Do you think evolution really cares whether constraints are artificial or not?

I knew someone was going to post this and should have addressed it..

Programming an evolutionary model is not setting artificial environmental constraints and then letting an unbounded process operate at will (as your comment implies). You would agree that the very act of coding a process puts constraints upon it. Programming any evolutionary model puts artificial constraints on both the environment and the evolutionary process itself.

Regardless, as "nsrivast" cleared up above, they are not striving to model Evolution (nor should they be claiming a process that has no constraints, intelligence or purpose is modeled accurately via constraints, a designer, and an end goal). As nsrivast noted, some of the most common constraints are time and computing power.

Re: Genetic Algorithm building a little car

#42

Post by bd »

ivankirigin wrote: post #40
bd wrote: post #35

Earlier quoted context omitted.

It's possible to get around using escape sequences. In Python: import urllib print urllib.quote("en.wikipedia.org/wiki/Smart_(automobile)") It looks ugly, but it makes links clickable: http://en.wikipedia.org/wiki/Smart_%28automobile%29

The href can be that ugly link while the parens are still visible. The issue is when to parse the parens as part of the link. Edit: but I guess you're saying I could do this to make it clickable. No thanks.

Yup, that's what I ment. This is what I do when posting Wikipedia urls with parens here.

For parsing parens problem, heuristics proposed by lacker seems reasonable.

Anyway, I guess like 99% of broken parens cases here would be Wikipedia links, which are quite well defined (except for pages for emoticons and parens symbols themselves).

Re: Genetic Algorithm building a little car

#43

Post by DavidSJ »

kirse wrote: post #24
andr wrote: post #5

Ok this is the second evolutionary computing post of today that does something cool with little or no explanation of what's going on. Stop teasing.

How exactly is it evolutionary computing when there is a designer in the equation (programmer) who puts constraints on the algorithm by the simple act of coding it and running it on systems that he designed? This seems more like "computing to model the process of genetic adaptation". You can't label it evolutionary when there's a designer in the equation. The very definition of the evolutionary process does not allow…

Your definition of evolution is incorrect. Evolution is a process which occurs in any system that has:

* replication

* selection pressure

* variation

No claims about that system's progeny are made by evolution.

Re: Genetic Algorithm building a little car

#45

Post by jwilliams »

vorador wrote: post #11

I wonder if the car that best fits the constraints is always the same (i.e : do we see the same car, or each user sees a different car) ?

I'd like to try and compete with an intelligent design - see how many generations it takes for the GA to beat me.

Re: Genetic Algorithm building a little car

#47

Post by anthonyb »

kirse wrote: post #41
anthonyb wrote: post #38

Earlier quoted context omitted.

Do you think evolution really cares whether constraints are artificial or not?

I knew someone was going to post this and should have addressed it.. Programming an evolutionary model is not setting artificial environmental constraints and then letting an unbounded process operate at will (as your comment implies). You would agree that the very act of coding a process puts constraints upon it. Programming any evolutionary model puts artificial constraints on both the environment and the evolution…

I think the quote that you're looking for is: "All models are wrong; some are useful".

That said, selection pressure is selection pressure, regardless of whether the means are artificial (being culled from a list in RAM) or natural (being eaten by a lion). I don't see that evolution in a computer process, even with artificial constraints, is that much different to evolution in the "real world".

I'm also not 100% sure exactly what your definition of an artificial constraint is, but something like Tierra might be more what you're looking for.

Re: Genetic Algorithm building a little car

#48

Post by wheels »

nsrivast wrote: post #29
kirse wrote: post #28

Earlier quoted context omitted.

If the programmer specifies the boundaries (laws) of his own universe, then the evolutionary processes that happen within his own universe are not representative of the actual universe which he claims to be modeling. Why? 1) Current scientific claims express that the laws and boundaries of the universe came into existence without a designer. Having a designer anywhere in the equation is unacceptable, even one who sim…

The point of genetic algorithms is NOT to replicate evolution. Like you mention, evolution has no end goal whereas engineers usually have very explicit end goals. (Other problems include time constraints and computing power!) The point of genetic algorithms is to exploit the principle of natural selection under very specific constraints and explicitly constructed utility functions that are likely to lead to a better…

To be clear, the field of genetic algorithms is split into two groups: one primarily concerned with function optimization for engineering application, the other at providing a laboratory for understanding evolutionary processes.

Re: Genetic Algorithm building a little car

#49

Post by demallien »

iamwil wrote: post #22
andr wrote: post #5

Ok this is the second evolutionary computing post of today that does something cool with little or no explanation of what's going on. Stop teasing.

What do you mean? I see the GA being tested against the fitness function one at a time with the individuals in the population, with subsequent populations being the mutated and crossover'd versions of the previous generation. Are you asking about the details of how each individual is represented? How the individuals are crossed over? What the rate of extinction is?

And I for one would love to know what the fitness function is, and what the physics being used is. And then, why not, what were the parametres that the ga was acting on...

Re: Genetic Algorithm building a little car

#50

Post by demallien »

nsrivast wrote: post #29
kirse wrote: post #28

Earlier quoted context omitted.

If the programmer specifies the boundaries (laws) of his own universe, then the evolutionary processes that happen within his own universe are not representative of the actual universe which he claims to be modeling. Why? 1) Current scientific claims express that the laws and boundaries of the universe came into existence without a designer. Having a designer anywhere in the equation is unacceptable, even one who sim…

The point of genetic algorithms is NOT to replicate evolution. Like you mention, evolution has no end goal whereas engineers usually have very explicit end goals. (Other problems include time constraints and computing power!) The point of genetic algorithms is to exploit the principle of natural selection under very specific constraints and explicitly constructed utility functions that are likely to lead to a better…

I beg to differ - evolution, in the 'life on Earth' sense has a very very specific goal - to reproduce genes as successfully as possible. To date, we know no life that does not seek this primal goal.
Post reply on HN