Name: OP !!tMEZoO6CDOz1DCF 2010-11-16 8:04
Sup /prog/,
I'm currently doing an AI hobby project and have hit a wall in progress due to my own stupidity.
I'm using Dijkstra's to search the tree. This of course requires Edge values, which, in this case, are decided by changes in state of Vertices [y=V(n)-V(n+1). Each specific change of state [x->x+1] being associated with an Edge. [y=Vn(x)-Vn+1(x+1)]
The specific issue I'm having is figuring out how much each node should be worth. I could arbitrarily assign values to each variable 'each x is worth y', but that feels rather naive. Would arbitrarily assigning values initially, and then using a heuristics to optimize on known test cases to assign 'better' values, be a good way of doing things?
Is there a better way of doing this?
The non-trivial nature of the tree means that an exhaustive search isn't possible, hence why I'm trying to optimize where I search as much as possible. This is going to be used to a) figure out what I should do and b) figure out what my opponent might do, so it's quite important that I get this algorithm spot on.
I'm currently doing an AI hobby project and have hit a wall in progress due to my own stupidity.
I'm using Dijkstra's to search the tree. This of course requires Edge values, which, in this case, are decided by changes in state of Vertices [y=V(n)-V(n+1). Each specific change of state [x->x+1] being associated with an Edge. [y=Vn(x)-Vn+1(x+1)]
The specific issue I'm having is figuring out how much each node should be worth. I could arbitrarily assign values to each variable 'each x is worth y', but that feels rather naive. Would arbitrarily assigning values initially, and then using a heuristics to optimize on known test cases to assign 'better' values, be a good way of doing things?
Is there a better way of doing this?
The non-trivial nature of the tree means that an exhaustive search isn't possible, hence why I'm trying to optimize where I search as much as possible. This is going to be used to a) figure out what I should do and b) figure out what my opponent might do, so it's quite important that I get this algorithm spot on.