Idea: object-based life-like CA simulator
Posted: July 31st, 2013, 12:49 pm
This is my idea, not yet an actual program.
Basically, the idea is a ca simulator focused on recognizing objects.
Before the program starts, it creates a hash and processes a table of common objects into a hash table.
When the user draws a pattern into the program, it does the following every turn:
First, it segments the pattern into individual objects, and looks up each pattern in a hash table.
For recognized objects, it pastes the known progression of the object into the next generation's array.
For unrecognized objects, it manually determines the progression of the object and enters it into the array, and if time and space permits, it sees the progression of the object and enters it and its rotations and reflections into the hash table.
This is probably slower than algorithms like hashlife, but it has the advantage of being able to recognize patterns more easily without reprocessing everything. It might help with pattern-finding: If we know a sacrificial eater coughs up a pi-heptomino that destroys it, we know where to look to stabilize it.
Ideas, tips, comments?
Basically, the idea is a ca simulator focused on recognizing objects.
Before the program starts, it creates a hash and processes a table of common objects into a hash table.
When the user draws a pattern into the program, it does the following every turn:
First, it segments the pattern into individual objects, and looks up each pattern in a hash table.
For recognized objects, it pastes the known progression of the object into the next generation's array.
For unrecognized objects, it manually determines the progression of the object and enters it into the array, and if time and space permits, it sees the progression of the object and enters it and its rotations and reflections into the hash table.
This is probably slower than algorithms like hashlife, but it has the advantage of being able to recognize patterns more easily without reprocessing everything. It might help with pattern-finding: If we know a sacrificial eater coughs up a pi-heptomino that destroys it, we know where to look to stabilize it.
Ideas, tips, comments?
My idea was that it links up with the Wiki and, when it sees a new object, it saves it to your "library" and the wiki. It also groups objects into sub-objects, and, creates a timeline graph of your pattern, showing which objects occur, how long they last and their fate. It's very hard to code these things because of tough cases such as guns and breeders, although it's plausible.