FRDB Archives

Freethought & Rationalism Archive

The archives are read only.


Go Back   FRDB Archives > Archives > IIDB ARCHIVE: 200X-2003, PD 2007 > IIDB Philosophical Forums (PRIOR TO JUN-2003)
Welcome, Peter Kirby.
You last visited: Today at 05:55 AM

 
 
Thread Tools Search this Thread
Old 07-23-2003, 11:06 PM   #1
Veteran Member
 
Join Date: Nov 2001
Location: NCSU
Posts: 5,853
Default Selection on Bitstrings

Here is an example of some of the modeling stuff I'm working on.

The Model
These are the results of a simulation I ran using bit-strings. Each bit-string contained 32 bits which represented a diallelic locus. The population consisted of 100x100 haploid, asexual cells. Constant viability selection was applied, such that an individuals fitness was determined by how similar its bit-string was to the optimal bit-string: w(i)=n(i)/32, where n(i) is the number of similarities. The probability that an individual dies in a generation is m(i)= m+(1-m)(1-w(i)), where m=0.01 is the background mortality rate. There was also a maximum age limit determined by the equation R=ln(0.05)/ln(1-m). When a cell died it was replaced by a mutated copy of a randomly picked neighbor. The expected number of mutations per replication was E(u)=1.

The Results
A full mp4 movie can be downloaded from here (1.7 mb). You might need Quicktime to view it. The color of an individual represents its fitness. (See scale bar: maroon=1, li green = 0.5, blue = 0.25, white = empty.)

What you can see in the simulation is that the population evolves from a collection of random bitstrings to a population of optimally fit indviduals in mutation-selection balance.





RufusAtticus is offline  
Old 07-24-2003, 05:47 AM   #2
Banned
 
Join Date: Mar 2001
Location: Boxing ring of HaShem, Jesus and Allah
Posts: 1,945
Default

What are you trying to demonstrate here, Rufus? This type of simulation - a mere collection of coloured dots - won't convince creationists of molecules-to-man evolution.

The ideal simulation is that which shows a complex machine arising out of no design at all.
emotional is offline  
Old 07-24-2003, 05:56 AM   #3
Contributor
 
Join Date: Sep 2000
Location: Alibi: ego ipse hinc extermino
Posts: 12,591
Default Re: Selection on Bitstrings

Quote:
Originally posted by RufusAtticus
What you can see in the simulation is that the population evolves from a collection of random bitstrings to a population of optimally fit indviduals in mutation-selection balance.
We can? Sorry Rufus, I'm sure that this is fascinating stuff... I don't want to belittle the probably huge amount of time you put into this... and maybe I'm missing something by only seeing the images (can't get the movie here at work)... but it just looks like a lot of pretty colour to me . My DarwinPond would appear to show evolution of improved fitness pretty well...?

Would you be kind enough to explain it all a bit more please? Till this thread, I was entirely happy not knowing what a bitstring even is... hells bells, is this something else I need to learn about?!

Cheers, Oolon
Oolon Colluphid is offline  
Old 07-24-2003, 05:57 AM   #4
Banned
 
Join Date: Sep 2001
Location: a place where i can list whatever location i want
Posts: 4,871
Default

Nothing's going to convince creationists. Especially not of "molocules to man evolution," whatever that's supposed to be.

I like your "mere colored dots," Rufus.
GunnerJ is offline  
Old 07-24-2003, 06:03 AM   #5
Banned
 
Join Date: Mar 2001
Location: Boxing ring of HaShem, Jesus and Allah
Posts: 1,945
Default

Quote:
Originally posted by GunnerJ
Nothing's going to convince creationists. Especially not of "molocules to man evolution," whatever that's supposed to be.
I beg to differ. Creationists usually complain of evolution that it is a "theoretical, historical science" and no-one was there to see evolution take place. And they often attack evolution by chugging out "information theory" - the question how information built up without design. I dare say an evolutionary computer simulation showing a complex machine emerging out of no design at all should shut all such objections up. Evolution would no longer be theoretical and historical once its creative power could be simulated on a computer screen.

Dots are just that - dots. Not to denigrate Rufus' hard work, but of much more value would be a simulation showing a complex cell arising out of front-loaded design (ie the parameters all set up, then running freely according to them).
emotional is offline  
Old 07-24-2003, 06:13 AM   #6
Regular Member
 
Join Date: May 2003
Location: US
Posts: 288
Default Re: Selection on Bitstrings

Nice post Rufus....but I have a few questions.....

Quote:
Originally posted by RufusAtticus
Here is an example of some of the modeling stuff I'm working on.

The Model
These are the results of a simulation I ran using bit-strings. Each bit-string contained 32 bits which represented a diallelic locus. The population consisted of 100x100 haploid, asexual cells. Constant viability selection was applied, such that an individuals fitness was determined by how similar its bit-string was to the optimal bit-string: w(i)=n(i)/32, where n(i) is the number of similarities. The probability that an individual dies in a generation is m(i)= m+(1-m)(1-w(i)), where m=0.01 is the background mortality rate. There was also a maximum age limit determined by the equation R=ln(0.05)/ln(1-m). When a cell died it was replaced by a mutated copy of a randomly picked neighbor. The expected number of mutations per replication was E(u)=1.
How was the optiaml bit-string of ( w(i)=n(i)/32 ) chosen? Where did the equation for m(i) come from? One last question........In the equation R=ln(0.05)/ln(1-m) is the "m" here the background mortality rate or is m the current value of m(i)?

Thanks.



Russ
Warcraft3 is offline  
Old 07-24-2003, 06:26 AM   #7
Regular Member
 
Join Date: May 2002
Location: Scotland
Posts: 155
Default Re: Re: Selection on Bitstrings

Quote:
Originally posted by Oolon Colluphid
We can? Sorry Rufus, I'm sure that this is fascinating stuff... I don't want to belittle the probably huge amount of time you put into this... and maybe I'm missing something by only seeing the images (can't get the movie here at work)... but it just looks like a lot of pretty colour to me . My DarwinPond would appear to show evolution of improved fitness pretty well...?

Would you be kind enough to explain it all a bit more please? Till this thread, I was entirely happy not knowing what a bitstring even is... hells bells, is this something else I need to learn about?!

Cheers, Oolon
What I think he has done is a simulation based on Goldberg Genetic Algorithm Agents??

Genetic Algorithms: In Search, Optimization and Machine Learning - David E Goldberg

Basically you have a bit string, a string of 1s and 0s, this is to simulate DNA "ADCG" etc, and then try to "evolve" the population to simulate a real life species weeding out the least fit.

So if it were 4 bit long strings:
1111 would be the "best" organism, most fittest, with a fitness of 15.
0001 would be weak with a fitness of 1.

if you had strings 1100 and 1010 and they crossed over at point three they may have offspring that are 1110 and 1000, add in random mutations and you try to reach the "goal" of fitness, in this case a maroon colour. The higher fitness normally only affects the chance to reproduce, so 1110 would have more offspring than 1000, leading to more offspring with a higher fitness, while the weak ones would not have as many offspring.

In this demonstration hes shown in a very basic way that starting with a random population, they gradually get stronger and stronger as a group with the weak being weeded out through lack of chance to reproduce.

P.s. how long did this take you, I have one I'm working on, have basically been doing a bit every few months, can never find enough time for it, I really should only concentrate on one programming project at a time
Alan G is offline  
Old 07-24-2003, 08:53 AM   #8
Veteran Member
 
Join Date: Jan 2001
Location: Santa Fe, NM
Posts: 2,362
Default

Not to rain too hard on Rufus's parade, but another valid criticism of this simulation is that the fitness function selected is so simple that the GA would be outclassed by a simple hill-climber (i.e., and optimzer that twiddles the bits in order and selects the locally optimal setting for that bit would get to optimial in 32 generations).

It might be interesting to try this with an N-K landscape or a SAT problem instead. These would have the advantage that the correct answer isn't obvious and preordained, and with multiple peaks, there's a reasonable chance of getting some speciation events going.
Undercurrent is offline  
Old 07-24-2003, 10:48 AM   #9
Veteran Member
 
Join Date: Nov 2001
Location: NCSU
Posts: 5,853
Default

Quote:
Originally posted by emotional
What are you trying to demonstrate here, Rufus?
That I can simulate evolution using bit-strings.

Quote:
This type of simulation - a mere collection of coloured dots - won't convince creationists of molecules-to-man evolution.
I don't care. My research goals are not dictated by what convinces a creationist.
RufusAtticus is offline  
Old 07-24-2003, 11:17 AM   #10
Veteran Member
 
Join Date: Jun 2001
Location: Denver, CO, USA
Posts: 9,747
Default

Quote:
Originally posted by emotional
I beg to differ. Creationists usually complain of evolution that it is a "theoretical, historical science" and no-one was there to see evolution take place. And they often attack evolution by chugging out "information theory" - the question how information built up without design. I dare say an evolutionary computer simulation showing a complex machine emerging out of no design at all should shut all such objections up. Evolution would no longer be theoretical and historical once its creative power could be simulated on a computer screen.
Such a thing would certainly not convince a creationist. First of all, having a complex machine arise "out of nothing" isn't exactly possible. You would have to simulate the environment in some fashion and begin with some sort of simple self-replicator. There are lots of evolutionary simulations that do this and they do come up with some highly complex critters. So your statement, "evolution would no longer be theoretical and historical once its creative power could be simulated on a computer screen," makes the false premise that this hasn't already been done, when in fact it has. But creationists simply say that this amounts to "intelligent design" because a human had to program in the environmental simulation, etc. Of course this is a lame response, and doesn't address what the programs are intended to show nor their utility, but it underscores the fact that a creationist will find something wrong with anything. They do the same thing with laboratory demonstrations of evolution -- the fact that a scientist set-up the experiment is taken to mean that the outcome was "designed". The goal-posts keep shifting. First they want you to simulate evolution. You do that, and then they want you to simulate it without setting up the conditions. Next they'll want you to simulate it without using any matter, because that would be cheating by using something that was already there and didn't evolve itself. If you think a simple experiment (or a series of them) is capable of convincing a creationist, you don't really know how they work.

theyeti
theyeti is offline  
 

Thread Tools Search this Thread
Search this Thread:

Advanced Search

Forum Jump


All times are GMT -8. The time now is 06:02 PM.

Top

This custom BB emulates vBulletin® Version 3.8.2
Copyright ©2000 - 2015, Jelsoft Enterprises Ltd.