Like Langton's Ant - Part 3

[TiagoTiagoT]

What if you check the patterns for each of the cells, count how many of each of the different patterns are present, the closer the numbers for all patterns are the better, perhaps with additional penalty for each pattern not present at all?

Also perhaps you could compare the pattern of each cell with its immediate neighbors, and add a penalty for each match; and perhaps also a compare with the patterns for additional layers around too, with the penalty for a match going down with distance.

[saken]

It would be interesting to use this for a stage in procedural texture generating. Maybe as a mask for splitting up other layers in various ways. Especially with an ability to control the rules over time as it's being generated, and perhaps blending between different rule sets.

[jarblewarble]

@saken It would be even better if you could generate Minecraft maps using a 3-dimensional rule of this type.

[saken]

@jarblewarble As part of the actual generating process I bet this would yield some interesting features. There is this Omen Map Editor that lets you create maps based on greyscale images, using the lightness/darkness to adjust the height. Usgs(dot)gov has loads of these DEMs (Digital Elevation Maps) for free. Which could potentially be used with it to generate some really interesting and original maps.

[TheSuperbloop]

Beats modern art.

[kjlg74]

@TheSuperbloop I should print a screen-capture image and sell it. I wouldn't be greedy or anything, a mere $10,000 should do it ;)

[jarblewarble]

@kjlg74 I created something like this once. I wonder if it could be used to create a map generator for some video games.

[ecreif]

another idea: calculate the amount of areas of the same color and reward the agent with the biggest and most "areas".

[ecreif]

let it try to paint a given picture or let it build a labyrinth by measuring the walkable space by using dijkstra :D

[kjlg74]

@ecreif Something like a reward for the largest contiguous "scribble" (for lack of a better word). Interesting idea.

[ecreif]

@kjlg74 yep shouldnt be too difficult to code it fast. am i right there are 8^512 possible agents? thats wayyyy more than atoms exist in the universe

[MaestroAlvis]

Second one doesn't look too bad. Couldn't really notice the reward for patterns. I think more specific patters need to be rewarded, having a certain squares be in certain colors. Also what if you tried larger configurations. I know considering the 8 serounding squares is kinda what this thing does but maybe giving rewards for patterns in the 8 serounding squares and then the 16 squares around that would create more visible results.

[kjlg74]

@MaestroAlvis Thanks for the suggestions. Perhaps I'll give it a try.

[Houshalter]

Since its tordial, when do you stop the program to determine fitness?

[kjlg74]

@Houshalter After a fixed number of iterations. Basically a time limit.

[TheReasonWhyGuy]

the wraping around the edge bit, removes the intensive to make dense complex figures :)

[kjlg74]

@TheReasonWhyGuy Yeah, that might be part of the trouble.

[BananaftRu]

most expensive way to get noise texture

[kjlg74]

@BananaftRu Yeah, lol. Not especially exciting.

[MrStripyHead]

thats my kind of ant farm!

[kjlg74]

@MrStripyHead That gives me an idea for my next tinkering-around-with-Java project! A virtual ant farm. Too many ideas, too little spare time.

[RoySchl]

@kjlg74 haha exactly what i thought when i read that comment.... "i could make an actual ant farm..." but yeah me "wanna do" and "should do" lists are both long and they dont get along well... i really need brain enhancement hardware in my head

[kjlg74]

@RoySchl I can definitely sympathize with that!