performance - Iterative deepening or trial and error? -
i coding board game. have generated game tree using alpha-beta pruning , have 2 options:
- use iterative deepening optimize alpha-beta keeps generating 1 more ply until time over.
- by trial , error, know maximum depth reachable every board configuration in time limit without inspecting lower plies.
which approach better , make search reach deeper depth? know, example, @ beginning can generate tree of depth x consuming time available... can iterative deepening add more depth?
let me know if can more clear...
if branching factor , evaluation function's required time stays same between different rounds might able use option 2. sounds might difficult.
option 1 has ability dramatically increase alpha-beta's performance if don't have perfect move ordering already. save best move @ every node, found during previous search d-1, , tell alpha-beta search first. it's called transposition table, , if possible implement in case option 1 dominate option 2.
the first time heard of transposition table didn't think make of difference, increased maximum search depth 50%.
Comments
Post a Comment