#haskell - Fri 23 Mar 2007 between 00:29 and 00:48

NY Lost Funds



chessguy@pl \f x y -> f (-y) (-x)
lambdabot(. subtract) . flip . (. subtract)
edwardkjust to avoid calculating the length of the list
Davemanwow, still at it chessguy? :)
chessguyDaveman: edwardk's doing the real work :)
edwardkheh
right now it works and is clean, just isn't necessarily as fast as i might want
chessguyi'm just wishing i knew what i was talking about
Davemanheehee :)
sounds interesting
edwardki think if i crack open the foldr and flip it around into a scanl, then i can muck around in its insides so it'll handle cases with infinite trees
right now its not entirely codata safe, because it foldrs
chessguyedwardk: what about the modifications we were talking about earlier?
mtd(f), etc.
edwardkwell, MTD(f) just uses a base alphabeta function and tweaks the alpha and beta parameters
iterative deepening just plays with the max depth
both of those are orthogonal to this implementation
chessguyand the original, really
edwardkiterative deepening would affect the game construction before it reaches here
MTD(f) tweaks the alpha beta values and binary searches for the right value setting alpha=beta
or beta = alpha + 1 depending on how its implemented
chessguyi'm not sure how much sense iterative deepening even makes in the presence of laziness
edwardkwell, you don't evaluate intermediate boards even while lazy
so iterative deepening is just as valuable here
i.e. i don't evaluate the value of a board 1 move away to get a better guesstimate of future board values with just laziness
ID still plays a role
also there is no memoization here, but that could be plugged in, probably monadically
chessguybut i mean, we could "evaluate" the "whole tree", but really only take the values to the depth that we want
edwardktrue, but there you'd have to change his game tree representation, right now his scoring function is baked into it. you can't evaluate intermediate levels
ID is just the ability to evaluate at progressively broader horizons.
chessguyright
oh yes, you'd want to change his tree structure to do really do any of this
besides which, using Data.Tree gives you unfoldTree, which is just too pretty not to use :)
fantasmawhat's a simple font viewer I can use from the console?
ddariusekidd: Okay, I don't think what I was thinking of is in Ghani's thesis. It might be. I'm having a hard time finding it.
edwardkheh
actionddarius should recompile his full-text database program.
ddariusshould recompile his full-text database program.
ekiddddarius: Well, if you run into it, please feel free to @tell me.
chessguyedwardk: well, thanks for the fun brain-melting session. it's time for bed here
desp?seen bos

Page: 2 9 16 23 30 37 44 51 58 65 72 79 86 93 100 107 114 

IrcArchive

NY Lost Funds