23 cells quadratic growth (new record)

For discussion of specific patterns or specific families of patterns, both newly-discovered and well-known.
mniemiec
Posts: 1055
Joined: June 1st, 2013, 12:00 am

Re: 23 cells quadratic growth (new record)

Post by mniemiec » July 19th, 2015, 5:07 pm

Gustavo6046 wrote:Seriously, you are Michael Simkin! I kept Michael Simkin in my ignore list! :shock: REMOVING NOW FROM IGNORE LIST

Oh wait, we are still rivals. I will also do a c/3 spaceship gun, other than dart, but this time the credits will be mine. And moreover, if it is to be a dart, my gun will be smaller and run at lower period! Unfortunately that means p30 with Gosper glider guns, but will hold a record!

When we get famous and scoreboards show 'Rehermann vs. Simkin' I will never admit it, but ONLY now I used your gun to make a p360 oscillator:
These forums are for discussing GOL and posting discoveries in it, not races to achieve personal glory at other people's expense. If you want glory, you're going to have to devote a lot of time and effort to come up with things that are totally unique, that nobody else has come up with, nor that was "obvious" from what went before. Your oscillator is trivially obvious (aiming guns at each other to annihilate their gliders is not new). Even factories to make exotic spaceships are fairly obvious (even though it may involve a fair bit of mechanical work), once the initial synthesis (NOT obvious) has been created. Assembling large arrays of guns and puffers, other than getting the timing and positioning right, is about difficult as assembling large objects out of Lego bricks. While the results may be nice to look at, they are hardly unique, nor worthy of glory.

Also remember that none of your discoveries would have been possible without major work done by others, which you then build on (e.g. the Gosper gun, eater eating glider, 3-glider collision to make a LWSS, Michael Simkin's gun, etc.) Painting a moustache on the Mona Lisa wouldn't suddenly make you the main artist and discard the work done by Leonardo da Vinci.

User avatar
simsim314
Posts: 1702
Joined: February 10th, 2014, 1:27 pm

Re: 23 cells quadratic growth (new record)

Post by simsim314 » July 20th, 2015, 3:54 am

mniemiec wrote:Assembling large arrays of guns and puffers, other than getting the timing and positioning right, is about difficult as assembling large objects out of Lego bricks.
I was thinking this way before I tried to construct something with really complex glider salvos. The problem is that the interaction that produces gliders has usually some smoke, and the smoke can collide with previous/next gliders. We don't have perfect inserter which is also minimal, so if we just add some constrain like - making it with smallest bounding box possible, the race becomes an open question that many smart people had hard time to improve. For example improving the dart gun bounding box, after dvgrn and cris_c worked hard to improve it, will be very impressive, at least it's more of an art than naive lego bricks.

The same goes for the Herschel conduits, making small and efficient construction using Herschels (like glider guns) is a very complex art that many people perfected to high extend, and it's everything but trivial lego assembling.

EDIT It's true that with some experience, assembling sketches of guns which are not optimal at all, or conduits is more like assembling lego bricks, but if someone wants to get to "production" level, something that would be close to optimal and useful for other projects, this is pretty complex area, that needs a lot of creativity, trickery, sometimes scripting and special interaction searches (for example one of the ways to fix the smoke issue is to have a glider to cancel the smoke, this requires extra special search - absolutely not lego).

mniemiec
Posts: 1055
Joined: June 1st, 2013, 12:00 am

Re: 23 cells quadratic growth (new record)

Post by mniemiec » July 20th, 2015, 6:01 am

simsim314 wrote:It's true that with some experience, assembling sketches of guns which are not optimal at all, or conduits is more like assembling lego bricks, but if someone wants to get to "production" level, something that would be close to optimal and useful for other projects, this is pretty complex area, that needs a lot of creativity, trickery, sometimes scripting and special interaction searches (for example one of the ways to fix the smoke issue is to have a glider to cancel the smoke, this requires extra special search - absolutely not lego).
Yes, this is what I meant. For example merely "making a crab gun" isn't difficult, once one has guns, salvo inserters, and a crab synthesis. "making a crab gun of small period X" or "within a bounding box of X*Y" CAN get progressively more challenging when one tightens the constraints. Then again, Lego is similar. Legos are subject to the laws of physics, so extremely large Lego projects tend to require some mechanical engineering and/or architectural skills.

User avatar
simsim314
Posts: 1702
Joined: February 10th, 2014, 1:27 pm

Re: 23 cells quadratic growth (new record)

Post by simsim314 » July 20th, 2015, 7:48 am

mniemiec wrote:extremely large Lego projects tend to require some mechanical engineering and/or architectural skills
I think when people refer to Lego they mean that it's simple straightforward "linear" assembling step by step task, nothing smart. Definitely Lego itself is bad example of this.

I was very surprised how complex it is to bring slow salvo recipes to production level. I mean one can split a block and then use some basic slow salvo recipes to construct, but then if SLs are placed too close, some of the recipes fail, then one needs not to split the block but use some trick to leave SL from the middle of previous recipe etc. I mean it becomes very complex and non linear task when you want to bring some "sketch" to really working, reasonably optimized prototype, that can be used for something more practical rather than just POC.

I never saw anything in GOL that was simple and straightforward based on recipes. There was always trickery and more complex issues involved other than what it seemed from the first glance. Recipes do work for most of the things most of the time, but you always need to improvise as well.

User avatar
gmc_nxtman
Posts: 1147
Joined: May 26th, 2015, 7:20 pm

Re: 23 cells quadratic growth (new record)

Post by gmc_nxtman » August 1st, 2015, 6:30 pm

Perhaps these ideas have been thought of before, but I'll just present them anyways. They may be able to decrease population count to 21-22 cells.

Idea 1:

A switch engine pair moves SE, producing SW glider streams. A glider is moving parallel to the switch engines. It catches up to the SW stream, and collides with one of the gliders. Then some of the other SW gliders collide with the debris. It makes a SW switch engine, and the reaction starts all over again. One would have to brute force Gliders colliding with side-shooting ark streams in order to find a working one if it exists.

Idea 2:

A Switch engine pair, once again, is moving SE, but has to produce backwards streams instead of side streams. A glider is, also, once again, moving SE parallel to them. However this time the glider is much farther behind. The backward gliders from the ark would normally pass harmlessly along the side of the puffers' debris, but the glider collides with it. It makes a growing crystal that periodically makes switch engines. One would also have to write a brute force program to collide gliders with backrake streams. This idea also could work if a stationary pre-block or blinker is placed close to the switch engine's origin. The backrake streams collide with the blonk, and causes a similar growing crystal.

Idea 3:

Another switch engine (pair) moves southeast. This one doesn't need to produce any gliders. A glider collides with the switch engine(pair)'s origin after several hundred ticks. The glider reacts with the debris produced by the Switch engine, and makes another SW (or NE for that matter) switch engine. It is also restored on the same path/color.

Idea 4:

A while back I found R-pentomino + Boat collision that made two perpendicular LWSS and MWSS. I lost it but it shouldn't be too hard to find again with gencols. Combining this with ten cells = Glider producing switch engine + MWSS/LWSS seed for Glider producing switch engine may make a smaller predecessor. Here's what I mean.

A NE Glider-streamer heads in its direction, meanwhile many cells to the NE of it the *WSS are being generated by the aforementioned collision. One of the *WSS continues onto being the "ping pong" part of the reaction, the other collides with a small selection of objects to produce the other glider producing switch engine. The result is a LWSS heading W, and two Glider-streamers heading NE, producing the Ping-pong reaction.

I hope these ideas are helpful.

User avatar
simsim314
Posts: 1702
Joined: February 10th, 2014, 1:27 pm

Re: 23 cells quadratic growth (new record)

Post by simsim314 » August 1st, 2015, 7:06 pm

Well without specific examples it's more of science fiction.

First of all before you make such design you need to have rough estimate on probability of SE to appear. The probability is somewhere between 1/100K and 1/M. So in order to give idea for search, you need to make sure the search space of valid options is of this order of magnitude.

1. There are about 100 different SE pairs, only some of them shoot non trivial glider streams. Having a glider stream you will probably get not more than 1K options to collide with it. Even in the best case scenario you will get one or two SE out of these collisions, but you need them to shoot exactly one extra glider in this 1K option space once more. So the search space is defined only by number of SE pairs (the more options the collision stream has, the harder it's to find exactly the right output to shoot again toward it). Making this idea lacking three orders of magnitude.

2. There are SE pairs shooting gliders backward, but good luck finding periodic crystals in there. The main problem with this approach, is that such crystals are very rarely periodic, thus the search space is ill defined. Other than that, the search space itself is very low, so you will be stuck with few thousand unclear crystals at best.

3. Once again the number of options to shoot extra glider, cancels out the coincidence that some output glider will shoot at the same spot once more. You again left with few hundred options.

---

Saying that, there is a chance that some "magic" will occur. I'm just saying magic is rare, and I wouldn't count on it - but it definitely happens once in a while - but you will be extremely lucky to get there basing on magic.

---

Other than that I've already posted a clear way to 22 cells without any risks at all, I was just busy with other stuff to complete it but 22 is not a big challenge, and hopefully with some work 21 will work as well but it requires to collect around 1K - 11 cells infinite growth glider producing SE, which is probably reachable because 8 = 5 (R) + 3 (blinker/block) which has around 100K options and adding another 3 in form of a blinker/block will add at least another 10K options, giving easily 1K glider producing SE.

User avatar
gmc_nxtman
Posts: 1147
Joined: May 26th, 2015, 7:20 pm

Re: 23 cells quadratic growth (new record)

Post by gmc_nxtman » August 1st, 2015, 7:44 pm

Did you use search programs to find those SE generators? If so, are you planning on releasing the program for finding small linear growth patterns?

User avatar
simsim314
Posts: 1702
Joined: February 10th, 2014, 1:27 pm

Re: 23 cells quadratic growth (new record)

Post by simsim314 » August 1st, 2015, 7:58 pm

gmc_nxtman wrote:Did you use search programs to find those SE generators? If so, are you planning on releasing the program for finding small linear growth patterns?
Oh I just write it from scratch, using golly python. It's about 20 lines of code so I don't bother to publish or even save it. I just make two steps in golly checking out the population grows linearly. Anyone with basic python skills can do it, and without python skills there is nothing to do with the code. I don't intend to make some user friendly "search tool".

The most sophisticated thing I did was a search for 6-7 cells constellations that lived for more than 800 ticks, to collide glider with them in hope of finding SE. Actually if there is some 10 cells I missed, is probably blinker/block collided with such 7 cells shorter living pattern.

User avatar
gmc_nxtman
Posts: 1147
Joined: May 26th, 2015, 7:20 pm

Re: 23 cells quadratic growth (new record)

Post by gmc_nxtman » August 1st, 2015, 8:59 pm

Here are some collisions you might want to try, of 10-11 cells. I definitely didn't put all of them, I just put easily constructible methuselah + easily constructible other object.

Code: Select all

B + Protoblonk
H + Protoblonk
Century + Tub
Century + Prepond
Century + PreTL
PreWing + Prepond
PreWing + Tub
PreWing + PreTL
LOM + Tub
LOM + PreTL
LOM + Prepond
Glider + R
Boat + R
R x 2
Pre-Pi + Boat
Pre-Pi + Glider
Pre-Pi + R
Pre-Pi x 2
Century + Tail

User avatar
simsim314
Posts: 1702
Joined: February 10th, 2014, 1:27 pm

Re: 23 cells quadratic growth (new record)

Post by simsim314 » August 2nd, 2015, 2:47 pm

gmc_nxtman wrote:Here are some collisions you might want to try, of 10-11 cells.
Well once again, before you try something you should roughly estimate the search space (number of different options you have). In general glider producing SE occurs once in 250K different options.

So lets measure the options you suggest:

LOM has trace of about 20x20, Century is 20x30, Wing is 25x20, Pi trace is about 40x50, R 50x150.


Now lets take Pi and R - we get 90 (50+40) x 200 (150 + 50) = 18000 options. Considering 8 orientations 144K options. So even taking the biggest 6 cells trace with huge 5 cells trace - we will get less than 250K options. I wouldn't say it's imposible to find there something, but for 11 cells to waste so much effort on such tiny search space is a real waste.

Now R+R and glider+R were searched pretty thoroughly and all the solutions are known, but to show the calculation for this case: R+R = 200x200x8 = 320K. There are also 3 different proto R (parent and grandparent), giving another factor of 3 (same, 1 tick dist, 2 ticks dist), in total we get ~M options or 4 glider producing SE (which is about right).

Glider + R is a bit more trickier calculation, as the glider can bring novelty as long as the pattern is evolving. Anyway I've found few such solutions last year and the search was exhaustive.

----

Now for 11 cells the only real thing that in the right search scope is R + protoblonk+protblonk. This gives the following estimation:

7500 (R) x 10000 (R + protoblonk tends to expand) x9 (protoblonk has 3 options) = 675M options. Or 2700 glider producing SEs. Compare this to all the options together in all your 11 cells suggestions (it was less than 1M).

Post Reply