Can HTM “learn” Conway’s Game of Life (in the sense of being able to predict the next board/configuration)?
(the problem probably gets easier if you do some game-specific pre-processing, but I’m imagining a scenario where that doesn’t happen)
Can HTM “learn” Conway’s Game of Life (in the sense of being able to predict the next board/configuration)?
(the problem probably gets easier if you do some game-specific pre-processing, but I’m imagining a scenario where that doesn’t happen)
It is a fair question. I can’t see why not, personally. Nice strict repeating state transitions.
The question is the HTM encoding - and I think this will be very specific to the game, since it doesn’t handle bitmaps well (in my limited experience). A locality approach should work well - especially if sized to match (more cheating ).
I’d certainly like to hear views from the community gurus.