##### Differences

This shows you the differences between two versions of the page.

 — cs-470:review-for-midterm [2015/01/06 14:51] (current)ryancha created 2015/01/06 14:51 ryancha created 2015/01/06 14:51 ryancha created Line 1: Line 1: + ==Agent Models== + *PEAS + *Types of Agents + *Types of Environments + *Decision Theoretic Model + + ==Search== + + I may use n-queens, checkers, chess, 8-puzzle (sliding block), rubic'​s cube, backgammon, tic-tac-toe as the context in any question on the exam. If you are not familiar with these, you may want to learn the basic idea of how they work. + + All algorithms '''​as described in the book'''​ + + Space complexity, Time complexity. You should be able to derive and discuss these. I prefer you not just memorize them. + + Optimality and Completeness. You should be able to prove and discuss these. + + You should know how all of the following algorithms work. This includes being able to simulate there execution for a small problem, and discuss their optimality, completeness,​ space complexity and time complexity. + + ===Uninformed=== + + Depth, Breadth, Uniform cost, Iterative Deepening, Bi-directional + + ===Informed search=== + + Greedy-best-first,​ A*, IDA*, Recursive Best-First Search, SMA* + + Heuristics, Admissibility,​ Consistency,​ Making Heuristics + + Tree vs. Graph search, Closed list. + + ===Beyond Classic Search=== + + On-line search + + Search in a continuous space + + Genetic Algorithms + + Simulated Annealing + + Particle swarm Optimization + + ==Games== + + Ply, Minimax, Terminal test, Evaluation functions, Cut-off, Quiescence search, Horizon problem, and Complexity + + Min/Max search + + $\alpha \beta$ pruning + + $\alpha \beta$ pruning w/ random nodes no limits, that is -$\infty$ to $\infty$ + + $\alpha \beta$ pruning w/ random nodes and limits + + == Probability == + + * Axioms of Prob. + * Definition of Conditional Prob. + * Notation, including: P(a) means the probability P(A=True), P(A) means a vector of probabilities corresponding to all values (all two, in the binary case) of the random variable A. + * Marginalizing out variable by summing, i.e. $p(a)=\Sigma_b P(a,b)$ + * Using the joint to compute arbitrary probabilities and arbitrary conditional probabilities + * Bayes Law + * Chain rule, using it both directions, that is to (1) split a joint probabilities '''​into'''​ conditionals and marginals, and (2) form joint probabilities '''​from'''​ conditionals and marginals. 