CSC 242 Lecture Notes - Lecture 3: Abstract Strategy Game, Iterative Deepening Depth-First Search
21 Feb 2017
School
Department
Course
Professor
Document Summary
Multi-agent environments: unpredictability of agents contingencies (strategies) (i. e. no single solution, agents goals in conflict competition. Games (side note: games require the ability to make some decision even when calculating the optimal decision is infeasible, games penalize inefficiency severely. Game problems are not toy problems as they have bigger state-spaces. Abstract games: deterministic (no chance, non-deterministic (cards, die, perfect information (fully observable, imperfect information (partially observable) (un-observable exists, zero-sum (total pay-off is constant) (not necessarily zero, arbitrary utility functions (in real life, not everything is zero-sum) Need to know: current state action terminal state current player depth - to get result faster. It would be a recursive case, getting the max utility on my turn, and min utility on your move, with terminal cases as base cases (book has code and examples). Instead of going for perfect decision (minimax), we can go for imperfect real-time decisions.
