The Online Life-Like CA Soup Search

For general discussion about Conway's Game of Life.
Post Reply
User avatar
Nathaniel
Site Admin
Posts: 624
Joined: December 10th, 2008, 3:48 pm
Location: New Brunswick, Canada
Contact:

The Online Life-Like CA Soup Search

Post by Nathaniel » June 29th, 2009, 5:30 pm

OK guys, it's "official" now: the soup search is up and running. See:

http://www.conwaylife.com/soup

This is now version 1.01. Here's what is new since the last version (v1.00) that I posted:
  • You can now view results for other rules (currently HighLife and 2x2 are the only non-standard rules officially supported). After I do some more testing I will unlock other rules or perhaps unlock all rules (including von Neumann neighbourhood rules).
  • The script not only searches for methuselahs, but also compiles a census. See this thread in the scripts forum for the Python source code of that census script. Census data is uploaded once every 15 minutes.
Here's what is coming in the near future (don't worry, these are changes that won't require a re-download of a new script version):
  • Categorization of objects by type (spaceship, still life, period, etc). Spaceships will be moved to a table separate from oscillators and still lifes, as it has been pointed out that the relative frequency of spaceships and stable/oscillating patterns depends heavily on the fact that the stats are generated by 20x20 starting patterns.
  • Animated images for oscillators/spaceships.
So go ahead and download the updated script! Your old (v1.00) scripts should be deactivated now, by the way.

Edit: Also, something I forgot to mention: unfortunately the census script is quite slow in HighLife, to the point of making about 60% of your search time dedicated to censusing. This is a result of HighLife's high tendancy to generate replicators, resulting in quite huge patterns before they actually stabilize (which is neat because it leads to extremely long-lived methuselahs, but makes the census script choke a bit). If you have read the census script I posted earlier and have insights about how to fix this problem, then please feel free to share. I'm still working on it.

Elithrion
Posts: 100
Joined: February 3rd, 2009, 4:02 pm
Contact:

Re: The Online Life-Like CA Soup Search

Post by Elithrion » June 29th, 2009, 5:40 pm

I imagine the navbar should say "Home", not "TOLLCASS Home" ;) Other than that, extracted and running okay so far.
Vi veri veniversum vivus vici.

Elithrion
Posts: 100
Joined: February 3rd, 2009, 4:02 pm
Contact:

Re: The Online Life-Like CA Soup Search

Post by Elithrion » June 29th, 2009, 10:56 pm

Hmm... the webpage didn't seem to update for a while while I was running 2x2. Not sure why.

On another note, the script was taking a while on this one time on HighLife, so I stopped it, restarted the pattern and decided to see what would happen. Well, I started it a little while ago. It's now at 1.9mil generations and now Golly is lagging and as far as I can tell it's not remotely stable. Here you go:

Code: Select all

x = 20, y = 20, rule = B36/S23
obo3bobo4bo$9bobo2bo2bobo$bob2o2b2o4bo2b2o$obo3bo3b2o2b4obo$ob8o2bo$2b
2ob2o2bo2b4obobo$b2obobobobo2bobob2o$2b3o2b2ob3ob2o3bo$3bo4b2o3bo$ob2o
3bo3b2o3bobo$bo2b2o2bob2o4bob2o$2bo2bo4bo4b2o2bo$7bo2bo2b3o2bo$bobo3bo
b2obo3bo2bo$2bo2b3obob3o4b2o$bo2bobobobob2obo3bo$obobo5bo2b5o$bo4bo2b
2ob2ob2o$o3bob2o3bobobo2b2o$4bo6bobo3b2o!
Vi veri veniversum vivus vici.

User avatar
Macbi
Posts: 744
Joined: March 29th, 2009, 4:58 am

Re: The Online Life-Like CA Soup Search

Post by Macbi » June 30th, 2009, 2:24 am

Elithrion wrote:Hmm... the webpage didn't seem to update for a while while I was running 2x2. Not sure why.

On another note, the script was taking a while on this one time on HighLife, so I stopped it, restarted the pattern and decided to see what would happen. Well, I started it a little while ago. It's now at 1.9mil generations and now Golly is lagging and as far as I can tell it's not remotely stable.
That's bizarre, maybe it never stabilises.

Elithrion
Posts: 100
Joined: February 3rd, 2009, 4:02 pm
Contact:

Re: The Online Life-Like CA Soup Search

Post by Elithrion » June 30th, 2009, 2:43 am

There is a problem with supposing that. Basically what's happening right now is that replicators are emerging sporadically (there are maybe some 12 or 13 alive right now), but they are also dying sporadically by turning into spaceships (there are maybe some 20-ish probably dead ones). While it's true that if this trend keeps up, there's going to be chaos forever, but at the same time, the probability that they will all just happen to die is not zero. There are some waves of gliders around that might help prevent that, but still, infinity is a long time and everything needs only to die once to never be alive again. That said, I would be downright shocked if it died before the 10 (or even 50) million generations mark. Also, here, a picture of what the beastie looks like right now (the sloping lines are waves of gliders, which separate replicators from debris; the long lines that aren't connected to the other stuff are replicators turned spaceships; and the little specks off to the sides are LWSSs).
Vi veri veniversum vivus vici.

User avatar
calcyman
Posts: 2226
Joined: June 1st, 2009, 4:32 pm

Re: The Online Life-Like CA Soup Search

Post by calcyman » June 30th, 2009, 10:46 am

The replicators are not turning into spaceships by the regular definition: they are aperiodic, and the head advances faster than the tail (ie. they expand). I'm not sure how good HashLife (Gosper's powerful algorithm, implemented by Rokicki for Golly) is at simulating these expanding XOR-ships. They aren't periodic, but they are more predictable than random junk.
Last edited by calcyman on July 3rd, 2009, 11:36 am, edited 2 times in total.
What do you do with ill crystallographers? Take them to the mono-clinic!

User avatar
Nathaniel
Site Admin
Posts: 624
Joined: December 10th, 2008, 3:48 pm
Location: New Brunswick, Canada
Contact:

Re: The Online Life-Like CA Soup Search

Post by Nathaniel » June 30th, 2009, 10:50 am

calcyman wrote:The replicators are not turning into spaceships by the regular definition: they areaperiodic, and the head advances faster than the tail (ie. they expand). I'm notsure how good HashLife (Gosper's powerful algorithm, implemented by Rokicki for Golly) is at simulating these expanding XOR-ships. They aren't periodic, but they are more predictable than random junk.
Unfortunately, it seems as though they're not periodic in a way that doesn't mesh well with HashLife -- it chugs almost as much as QuickLife does.

I imagine that I should remove the "Longest-Lived" table for HighLife altogether, since it's clear that methuselahs behave a bit differently in the world of B36/S23 than they do in the usual setting... and I'll throw in some sort of trigger for the next version that stops HighLife patterns from evolving past 30000 generations, or something like that.

Elithrion
Posts: 100
Joined: February 3rd, 2009, 4:02 pm
Contact:

Re: The Online Life-Like CA Soup Search

Post by Elithrion » June 30th, 2009, 1:50 pm

Yeah, that's probably a good idea. I let my pattern run overnight (my internet dies overnight for some reason anyway), and it hit 2.5mil generations. As expected, it's more of a mess than ever. I'll go do something productive now, but if someone wants the RLE of that generation for some reason, let me know (although I doubt anyone does :)).
Vi veri veniversum vivus vici.

User avatar
Macbi
Posts: 744
Joined: March 29th, 2009, 4:58 am

Re: The Online Life-Like CA Soup Search

Post by Macbi » June 30th, 2009, 2:26 pm

Elithrion wrote:Yeah, that's probably a good idea. I let my pattern run overnight (my internet dies overnight for some reason anyway), and it hit 2.5mil generations. As expected, it's more of a mess than ever. I'll go do something productive now, but if someone wants the RLE of that generation for some reason, let me know (although I doubt anyone does :)).
That looks like quadratic growth, which would seem to imply that every sufficiently large highlife pattern tends to explode quadratically, which is unexpected.

Elithrion
Posts: 100
Joined: February 3rd, 2009, 4:02 pm
Contact:

Re: The Online Life-Like CA Soup Search

Post by Elithrion » June 30th, 2009, 3:06 pm

Well, it definitely looks like quadratic growth over that stretch, but I'm not sure it would be sustained. It's very very chaotic quadratic growth at best. I mean, it's sort of explainable. Say, if you start with a replicator and some debris, there's some probability that it will create another replicator going sideways from it (which happened with my pattern on several occasions), and also some chance that it will turn into a pseudo-spaceship-thingie (what would you call that exactly?) If perpendicular replicators are made often enough, then quadratic growth is assured! More or less.

On another note, Nath, the replicator seems to have gotten census-ed (rank 36 just now)! I'm not entirely sure what to make of that.
Vi veri veniversum vivus vici.

User avatar
Lewis
Posts: 332
Joined: March 17th, 2009, 5:26 pm
Location: UK
Contact:

Re: The Online Life-Like CA Soup Search

Post by Lewis » June 30th, 2009, 3:22 pm

Will the rule B35/S23 ever be supported by the software? A while ago I did some research into it and found some interesting oscillators (2 P8's, a few P4's and a 2c/5 spaceship which all occur naturally.)

User avatar
Nathaniel
Site Admin
Posts: 624
Joined: December 10th, 2008, 3:48 pm
Location: New Brunswick, Canada
Contact:

Re: The Online Life-Like CA Soup Search

Post by Nathaniel » June 30th, 2009, 3:50 pm

Elithrion wrote:On another note, Nath, the replicator seems to have gotten census-ed (rank 36 just now)! I'm not entirely sure what to make of that.
Thanks for the notice, I've added that version of it to the "don't catalog me" list.
Lewis wrote:Will the rule B35/S23 ever be supported by the software? A while ago I did some research into it and found some interesting oscillators (2 P8's, a few P4's and a 2c/5 spaceship which all occur naturally.)
Sure thing. Basically just request a rule and after making sure that it's "censusable" (ie. not explosive) I'll add it to the list (when I get time ;)). I'll keep B35/S23 in mind for the near future (I also want to add Pseudo Life and a handful of others).

User avatar
Extrementhusiast
Posts: 1837
Joined: June 16th, 2009, 11:24 pm
Location: USA

Re: The Online Life-Like CA Soup Search

Post by Extrementhusiast » June 30th, 2009, 4:03 pm

B3/S015678.
I Like My Heisenburps! (and others)

User avatar
Lewis
Posts: 332
Joined: March 17th, 2009, 5:26 pm
Location: UK
Contact:

Re: The Online Life-Like CA Soup Search

Post by Lewis » June 30th, 2009, 4:16 pm

Some more Rule requests:

Day & Night
B24/S13 in the Von Neumann neighborhood (if Golly can run it)
b35678/s248 (has interesting high-period oscillators)
B3/S24
B35/S23

User avatar
Nathaniel
Site Admin
Posts: 624
Joined: December 10th, 2008, 3:48 pm
Location: New Brunswick, Canada
Contact:

Re: The Online Life-Like CA Soup Search

Post by Nathaniel » June 30th, 2009, 6:19 pm

I just added Day & Night, which works quite well. I'll get around to the others in the near future, interspersed with other updates.

Elithrion
Posts: 100
Joined: February 3rd, 2009, 4:02 pm
Contact:

Re: The Online Life-Like CA Soup Search

Post by Elithrion » June 30th, 2009, 8:13 pm

So, in 850k soups of Day & Night, we've apparently found one "long-lived" pattern it. Turns out it's a spaceship.

[edit:] on another note the RLE's aren't being output correctly on the webpage - the bounding boxes are off. So, for example, the one for "rocket" (incidentally, where'd you pull that name from?) says x=20,y=20, even though it's something like x=46,y=9.
Vi veri veniversum vivus vici.

User avatar
Nathaniel
Site Admin
Posts: 624
Joined: December 10th, 2008, 3:48 pm
Location: New Brunswick, Canada
Contact:

Re: The Online Life-Like CA Soup Search

Post by Nathaniel » July 1st, 2009, 3:25 am

Elithrion wrote:So, in 850k soups of Day & Night, we've apparently found one "long-lived" pattern it. Turns out it's a spaceship.

[edit:] on another note the RLE's aren't being output correctly on the webpage - the bounding boxes are off. So, for example, the one for "rocket" (incidentally, where'd you pull that name from?) says x=20,y=20, even though it's something like x=46,y=9.
Yeah, I'll have to lower to default limit for Day & Night (it's set to 2500 generations before any long-lived patterns are found for that rule). I'll try to get that bounding box thing fixed shortly too... as for "rocket" -- it comes from David Eppstein's "Life-like glider" page (here).

Also, found a neat wickstretcher in Day & Night, which has almost surely been found before:

Code: Select all

x = 31, y = 17, rule = B3678/S34678
3$7bo$6bobo2b2obo6bo$5bob2o2b6o3b4o$6b11o2b6o$5b21o$6b21o$6b21o$4b22o$
5b5ob9o4bo$6b2o5b3ob4o2bo!

Elithrion
Posts: 100
Joined: February 3rd, 2009, 4:02 pm
Contact:

Re: The Online Life-Like CA Soup Search

Post by Elithrion » July 1st, 2009, 7:44 pm

Hmm... your rotation/reflection check seems not like dealing with very many objects or something. Or maybe just with Move. Note for example the rank 10-17 objects which are just obvious rotations/reflections of each other. Also note the big puffers at present ranked 288/312/318/335 and 300/305/316/334 respectively. Now also note that those are, in fact, the same (p170) puffer ;)
[edit:] okay, not just Move - the same thing seems to be happening with 2x2 as well to some extent. Did you change something today that might have screwed the check up, perhaps?

On another note, I like what you've done with the oscillator pictures and the frequency counts, by the way. Although now I must ask - what confidence interval is that, 95%?
Vi veri veniversum vivus vici.

User avatar
Nathaniel
Site Admin
Posts: 624
Joined: December 10th, 2008, 3:48 pm
Location: New Brunswick, Canada
Contact:

Re: The Online Life-Like CA Soup Search

Post by Nathaniel » July 2nd, 2009, 2:43 am

Elithrion wrote:Hmm... your rotation/reflection check seems not like dealing with very many objects or something. Or maybe just with Move. Note for example the rank 10-17 objects which are just obvious rotations/reflections of each other.
There's some basic clean-up that I have to do on the data, and I haven't started on Move yet (my project yesterday was the confidence intervals and the ability to add oscillator images/periods ;)). I'll get caught up on Move and 2x2 today (basically I run a script on my end of thing that merges relevant rows together).
Elithrion wrote:On another note, I like what you've done with the oscillator pictures and the frequency counts, by the way. Although now I must ask - what confidence interval is that, 95%?
It's 99%. I'll add that info to the FAQ once I get around to writing it.

User avatar
Lewis
Posts: 332
Joined: March 17th, 2009, 5:26 pm
Location: UK
Contact:

Re: The Online Life-Like CA Soup Search

Post by Lewis » July 2nd, 2009, 2:25 pm

Why does the results for each rule only go down to patterns that have occured 10 or more times?

Elithrion
Posts: 100
Joined: February 3rd, 2009, 4:02 pm
Contact:

Re: The Online Life-Like CA Soup Search

Post by Elithrion » July 2nd, 2009, 2:57 pm

Nath mentioned at one point that it was an arbitrary limit he set (and that it used to be 25). Presumably below that the frequencies/rankings just become pretty arbitrary, statistically speaking. Although I guess with the confidence intervals in place, it might be okay to just deal with all the objects, maybe? Of course it would be better to have the rankings be a range as well or something.
Vi veri veniversum vivus vici.

User avatar
Nathaniel
Site Admin
Posts: 624
Joined: December 10th, 2008, 3:48 pm
Location: New Brunswick, Canada
Contact:

Re: The Online Life-Like CA Soup Search

Post by Nathaniel » July 2nd, 2009, 3:34 pm

Lewis wrote:Why does the results for each rule only go down to patterns that have occured 10 or more times?
In addition to what Elithrion said, there's some minor clean-up I have to do yet with the rankings before they make sense at low counts (see the bottom of the 2x2 page right now to see what I mean -- they're more or less gibberish). This will be better once I get reliable internet (in about a week) and can actually do things online for more than 10 minutes at a time.

User avatar
calcyman
Posts: 2226
Joined: June 1st, 2009, 4:32 pm

Re: The Online Life-Like CA Soup Search

Post by calcyman » July 3rd, 2009, 11:09 am

Why does the results for each rule only go down to patterns that have occured 10 or more times?

So that the information is reliable. It's possible that a freak accident causes a really complicated still-life to emerge naturally, but setting a limit helps to eliminate this possibility.
What do you do with ill crystallographers? Take them to the mono-clinic!

User avatar
Lewis
Posts: 332
Joined: March 17th, 2009, 5:26 pm
Location: UK
Contact:

Re: The Online Life-Like CA Soup Search

Post by Lewis » July 8th, 2009, 2:07 pm

Where did the census pages for the other rules go? Only the census for the stardard rules shows for me.

User avatar
Nathaniel
Site Admin
Posts: 624
Joined: December 10th, 2008, 3:48 pm
Location: New Brunswick, Canada
Contact:

Re: The Online Life-Like CA Soup Search

Post by Nathaniel » July 8th, 2009, 2:14 pm

Lewis wrote:Where did the census pages for the other rules go? Only the census for the stardard rules shows for me.
Sorry, I'm doing some major changes this week. Things will be back up and normal (with new features, of course) as of Friday or Saturday.

Post Reply