flat assembler
Message board for the users of flat assembler.

Index > Heap > Another time waster...

Goto page 1, 2  Next
Author
Thread Post new topic Reply to topic
bitRAKE



Joined: 21 Jul 2003
Posts: 2906
Location: [RSP+8*5]
bitRAKE

_________________
¯\(°_o)/¯ unlicense.org
Post 25 Dec 2008, 03:59
View user's profile Send private message Visit poster's website Reply with quote
vid
Verbosity in development


Joined: 05 Sep 2003
Posts: 7105
Location: Slovakia
vid
cute Smile
Post 25 Dec 2008, 13:51
View user's profile Send private message Visit poster's website AIM Address MSN Messenger ICQ Number Reply with quote
Borsuc



Joined: 29 Dec 2005
Posts: 2466
Location: Bucharest, Romania
Borsuc
I think even Brute-Force is faster than this Razz
Post 25 Dec 2008, 14:54
View user's profile Send private message Reply with quote
vid
Verbosity in development


Joined: 05 Sep 2003
Posts: 7105
Location: Slovakia
vid
Why do you think so?

AFAIK bruteforce tries all possibilities, while evolution tries possibilities closely derived from subset that perform task best. Thus, evolution is much more likely to eliminate completely bad possibilities, which still must be tested by brute force.
Post 25 Dec 2008, 15:50
View user's profile Send private message Visit poster's website AIM Address MSN Messenger ICQ Number Reply with quote
edfed



Joined: 20 Feb 2006
Posts: 4237
Location: 2018
edfed
interresting.
i guess what is the approach to code something like that.
Post 25 Dec 2008, 16:38
View user's profile Send private message Visit poster's website Reply with quote
Borsuc



Joined: 29 Dec 2005
Posts: 2466
Location: Bucharest, Romania
Borsuc
Of course I'm not talking about mindless brute force, but about some form of "backtracing". I've seen the car doing "great" and then "evolving" and doing poorly... or maybe devolving. This doesn't happen in backtracing, it never 'devolves' Wink

_________________
Previously known as The_Grey_Beast
Post 25 Dec 2008, 16:39
View user's profile Send private message Reply with quote
bitRAKE



Joined: 21 Jul 2003
Posts: 2906
Location: [RSP+8*5]
bitRAKE
'devolving' is important to escape local minima.

Some have evolved to getting past the second hill after the double drop. Most seem springy to regain energy from a drop. Unfortunately, the graph is relative and can't be compared across separate executions.

Initial success (or lack of) seems to have a huge impact. For example, the quad finally changed to triangles, but reverts back to quads. Yet none of the triangle starters are trying quads it seems. {edit} Recently, had two successful quads in the first generation (how lucky?), and all following generations have been quads -- will be interesting to see how far it gets.

Bet ya can't guess which one gets furthest (trick question):


Description: A selection of several hour old cars.
Filesize: 5.74 KB
Viewed: 4835 Time(s)

2DCars.png


Post 25 Dec 2008, 19:22
View user's profile Send private message Visit poster's website Reply with quote
bitRAKE



Joined: 21 Jul 2003
Posts: 2906
Location: [RSP+8*5]
bitRAKE
edfed wrote:
interresting.
i guess what is the approach to code something like that.
http://www.custom-logic.com/exp/cloth/cloth.html
Very Happy
I kind of want to code my own, and try other configurations!

_________________
¯\(°_o)/¯ unlicense.org
Post 01 Jan 2009, 03:25
View user's profile Send private message Visit poster's website Reply with quote
revolution
When all else fails, read the source


Joined: 24 Aug 2004
Posts: 17253
Location: In your JS exploiting you and your system
revolution
vid wrote:
AFAIK bruteforce tries all possibilities, while evolution tries possibilities closely derived from subset that perform task best. Thus, evolution is much more likely to eliminate completely bad possibilities, which still must be tested by brute force.
Evolution can't solve the local maximum problem.
bitRAKE wrote:
'devolving' is important to escape local minima.
Just to further amplify on bitRAKE's comment.

Suppose you are trying to climb to the highest point in your domain. You have limited vision and can only see one step ahead of yourself. So at the start you find yourself on a slope and naturally you want to take a step "up" that slope in order to be at a higher position. You continue to do this and kind of wander around the hill for some time until finally you reach the top and find that all further steps lead downwards. So you conclude that you have reached the highest point. But, what you don't realise, because of the limited vision, is that there is a mountain on the other side of the valley that you never found. If you had only taken a few steps "down" from your initial position at the start then you would have discovered another rising slope to be explored.
Post 01 Jan 2009, 03:50
View user's profile Send private message Visit poster's website Reply with quote
vid
Verbosity in development


Joined: 05 Sep 2003
Posts: 7105
Location: Slovakia
vid
Quote:
Evolution can't solve the local maximum problem.

I think it can, to certain extent given by level of toleration (less strict selection). Of course, too much would risk devolving instead of evolving.
Post 01 Jan 2009, 13:12
View user's profile Send private message Visit poster's website AIM Address MSN Messenger ICQ Number Reply with quote
revolution
When all else fails, read the source


Joined: 24 Aug 2004
Posts: 17253
Location: In your JS exploiting you and your system
revolution
But consider what happens when you are at the top of the local hill (my example above), you need to do a lot of devolving (down steps) before you can start to see any benefit coming. Once you are at the top of that hill it is very hard to find the mountain. Evolution has no direction to follow, it doesn't have a future plan to find mountains. The is no guiding directive "keep going South regardless of the terrain". If something starts to look like a disadvantage it has a strong tendency to be eliminated.
Post 01 Jan 2009, 13:43
View user's profile Send private message Visit poster's website Reply with quote
vid
Verbosity in development


Joined: 05 Sep 2003
Posts: 7105
Location: Slovakia
vid
Yeah, of course i agree about that, but I said that evolution can overgo local maxima to certain extent - unlike what Borsuc suggested.
Post 01 Jan 2009, 14:51
View user's profile Send private message Visit poster's website AIM Address MSN Messenger ICQ Number Reply with quote
bitRAKE



Joined: 21 Jul 2003
Posts: 2906
Location: [RSP+8*5]
bitRAKE
The dimensionality precludes exhaustive search. Why shouldn't exhaustive search be part of evolution (operating in some sub-set of the problem domain, of course)? The environment puts constraints on the problem which can reduce evolution to a tractable sub-set. How to climb all hills at once!? The problem is incorrectly modeling the environment - the hills are not stationary and the climber dies (top of a hill or not).
Post 01 Jan 2009, 17:08
View user's profile Send private message Visit poster's website Reply with quote
Borsuc



Joined: 29 Dec 2005
Posts: 2466
Location: Bucharest, Romania
Borsuc
Because evolution doesn't treat all options "the same" (with equal probability). So like revolution said, it can eventually be eliminated. Brute force says "heck it", and tries them, when evolution thinks "nah, it's going to be bad" so it assigns less probabilities so to speak. The cool thing about Brute Force is that if it happens to be bad, it simply goes back (backtracing) to what it was before, and realizes it was bad, without any kind of modification and starts another path. Like making an experiment that blows up your house -- I think it's more convenient to "return" back WITH your house trying a different experiment (because that one went wrong) than to "deal with" it without any house, and continue like that.


A neural net would work better I guess. (that is, not a primitive 'evolution' style, but a 'planning' style Razz; after it makes a mistake, it plans about the accumulated information)
Post 01 Jan 2009, 19:33
View user's profile Send private message Reply with quote
MSWarrior



Joined: 21 Mar 2008
Posts: 11
MSWarrior
Ever wondered how such a creature will look in the real world?
Well take a look at these strange beings. They just might be the dinosaurs of the 21st century Smile
Post 22 Jan 2009, 19:58
View user's profile Send private message Reply with quote
HyperVista



Joined: 18 Apr 2005
Posts: 691
Location: Virginia, USA
HyperVista
Shocked
Post 22 Jan 2009, 20:15
View user's profile Send private message Visit poster's website Reply with quote
Borsuc



Joined: 29 Dec 2005
Posts: 2466
Location: Bucharest, Romania
Borsuc
I'm speechless.

I heard of cyborgs. They don't impress me. But this thing is 100% mechanical Shocked

_________________
Previously known as The_Grey_Beast
Post 22 Jan 2009, 21:28
View user's profile Send private message Reply with quote
bitRAKE



Joined: 21 Jul 2003
Posts: 2906
Location: [RSP+8*5]
bitRAKE
A common question is how the legs really work in Theo Jansen's designs - like if someone wanted to build such a thing themselves. Since I found this I'll save others the search. (Theo used a genetic algorithm to evolve the lengths.)

http://www.mechanisms101.com/theo_jansen.shtml
Post 23 Jan 2009, 05:39
View user's profile Send private message Visit poster's website Reply with quote
vid
Verbosity in development


Joined: 05 Sep 2003
Posts: 7105
Location: Slovakia
vid
Quote:
They just might be the dinosaurs of the 21st century

Strictly speaking, all birds are dinosaurs, so there is nothing special about "dinosaurs of 21st century"...
Post 23 Jan 2009, 12:44
View user's profile Send private message Visit poster's website AIM Address MSN Messenger ICQ Number Reply with quote
revolution
When all else fails, read the source


Joined: 24 Aug 2004
Posts: 17253
Location: In your JS exploiting you and your system
revolution
vid wrote:
Strictly speaking, all birds are dinosaurs, ...
As proved by Pete.
Post 23 Jan 2009, 13:28
View user's profile Send private message Visit poster's website Reply with quote
Display posts from previous:
Post new topic Reply to topic

Jump to:  
Goto page 1, 2  Next

< Last Thread | Next Thread >
Forum Rules:
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum
You can attach files in this forum
You can download files in this forum


Copyright © 1999-2020, Tomasz Grysztar.

Powered by rwasa.