1
   

Exponential probability

 
 
stuh505
 
Reply Sat 8 Oct, 2005 05:56 pm
Here's a problem which seems pretty common in nature...but I cannot think of any conventional way to write an equation for it!!!

You start with 1 node. At each time epoch, each node has a chance to split, creating 2 new nodes. Otherwise it creates 1 new node. With an arbitrary split probability (0.0 - 1.0), how can an equation be written?

If the split probability were 0, the equation would be:

nodes = time

If the split probababiltiy were 1 (100%), the equation would be:

nodes = 2^time

But what if the probability is 0.5?
  • Topic Stats
  • Top Replies
  • Link to this Topic
Type: Discussion • Score: 1 • Views: 621 • Replies: 5
No top replies

 
markr
 
  1  
Reply Sat 8 Oct, 2005 07:07 pm
If at each time epoch each node creates one new node (split probability = 0), won't nodes = 2^time?

For 0 < P < 1, f(t, 0) <= f(t, P) <= f(t, 1)

I don't see how you can come up with an exact formula unless you assume that at each time epoch exactly P*100% of the nodes split. In that case, you'll likely end up with a non-integer number of nodes.
0 Replies
 
engineer
 
  1  
Reply Sat 8 Oct, 2005 07:45 pm
If P is a fixed value for the entire tree, then I think the value would approach (1+P)^time. If P is variable for each split, then 1.5^time. If you repeated this numerous times, my guess is that the standard deviation of ln(nodes) would be [sqrt( time * P * (1-P) )] ln(1+P). I'll have to put it in excel and run a few hundred cases to test that out.
0 Replies
 
stuh505
 
  1  
Reply Sun 9 Oct, 2005 12:00 am
Quote:
If at each time epoch each node creates one new node (split probability = 0), won't nodes = 2^time?


Yes, but only because I didn't describe it right Sad
Only the most recently created nodes split/generate new nodes.

So for example...if split probability = 0, then we have this:

total nodes:
1, 2, 3, 4, 5...

new nodes:
1, 1, 1, 1, 1

If split probability =1, we have this:

total nodes:
1, 3, 7, 15, ...

new nodes:
1, 2, 4, 8, ...

If split probabiltiy = 0.5, the amount of new nodes per turn could follow any of these paths:

http://img28.imageshack.us/img28/2028/split3kg.jpg

Quote:
If P is a fixed value for the entire tree, then I think the value would approach (1+P)^time


P is fixed for the entire tree but with this equation if P=0 then you are left with total nodes = 1 always...

I am not looking for an exact answer, just looking for an approximate or the lowest possible upper bound
0 Replies
 
markr
 
  1  
Reply Sun 9 Oct, 2005 01:43 am
Do you really mean to have three branches from a single node?

Isn't the lowest possible upper bound the same as when P=1? Of course, the probability of realizing this upper bound approaches zero (if P < 1) as t approaches infinity.

I think the expected value of f(P,t) = SUM(i=0 to t, (1+P)^i)

I think engineer was just counting terminal nodes.
0 Replies
 
stuh505
 
  1  
Reply Sun 9 Oct, 2005 08:39 am
Ah ok...that equation looks pretty good then.

What if there was another probability P2 that ANY previous node will create 1 new node?

The nodes in the tree that I drew are not the same as the nodes that I am counting...the nodes I was drawing represent the state of the change, so if I have 2 nodes created last epoch then during the next epoch both nodes could not split, creating 2 new nodes, or both nodes could split, creating 4 new nodes, or only 1 of them could split, creating 3 nodes. so, three branches for 3 options.
0 Replies
 
 

Related Topics

Evolution 101 - Discussion by gungasnake
Typing Equations on a PC - Discussion by Brandon9000
The Future of Artificial Intelligence - Discussion by Brandon9000
The well known Mind vs Brain. - Discussion by crayon851
Scientists Offer Proof of 'Dark Matter' - Discussion by oralloy
Blue Saturn - Discussion by oralloy
Bald Eagle-DDT Myth Still Flying High - Discussion by gungasnake
DDT: A Weapon of Mass Survival - Discussion by gungasnake
 
  1. Forums
  2. » Exponential probability
Copyright © 2024 MadLab, LLC :: Terms of Service :: Privacy Policy :: Page generated in 0.04 seconds on 05/01/2024 at 03:05:49