Topics

• Problem Spaces and Defintions
• Search
• Complete / Optimal
• Uninformed -- DFS / BFS / UCS
• Informed Seaches
• • Heuristics -- Admissible, Monotonic
  • A*, Greedy
  • Iterative searches: IDA*, IDS
• Local Search
• Hill Climbing
• Simulated Annealing
• Genetic Algorithms
• Bitstring Fitness
• Crossover
• Mutation
• Selecting Parents for Crossover: Roulette Selection vs Tournament Selection
• Eliteism
• Adversarial Search
• Min-Max
• Alpha-Beta Pruning
• Evaluation Functions
• Propositional Logic
• Converting sentences into propositional logic
• Converting into CNF
• Resolution and refutation

Potential Questions

1. Definition Questions. True / False / Explain. If it is true, mark it true. If it is false, correct the statement so that it is true. Note: Adding ``not'' or otherwise negating the sentence is not acceptable. You must change the facts in the sentence if it is false.

   Question: The Turing test is a test of whether a computer program is rational

   Bad Answer, no credit: The Turing test is {\em not} a test of whether a computer program is rational

   Good Answer: The Turing Test is a test of whether a computer program is indistinguishable from a human.

   1. A complete search algorithm is one that is guaranteed to always find an optimal solution
   2. A stochastic environment is one in which the world does not change when the agent is deciding upon an action
2. Problem Setup. Given a problem defintion, draw a graph of the state space, with initial state and goal states specified, and all operations which connect states to other states.
3. Search Execution. Given a partial search space with the heursitic value of each node and cost of each operator labeled
1. Determine if the heuristic function is definitely admissible, definitely not admissible, or possibly admisible
2. Give the order in which nodes are expanded (not generated) under A*
3. Give the order in which nodes are expanded under IDA* (be sure to include all iterations, so that some nodes will appear in your list several times. Assume that on each iteration, the new depth cutoff is set to the f-value of the largest unexpanded node)
4. Given a 2-player game tree, with values at the leaves
1. Back up the values from leaves to the root
2. Give the sequence of moves that will be made if both players are rational
3. Assuming the min-max search traverses the tree from left to right, circle the nodes that will not be examined if alpha-beta pruning is used.
5. Genetic Algorithms. Given an encoding of a problem and a fitness function:
1. For a specific population, give the fitness for each element, and the problability that each element would be chosen by roulette selection
2. Show the resulting children for a given crossover, given the parents and the crossover point
3. What makes a good encoding? What makes a bad encoding?
6. Logic
1. Given a set of propositions, convert English senteces into propositional logic
2. Convert into CNF
3. Use resolution by refutation to prove a specific proposition is true
Example: Using the following propositions:
• S: Smith wins the election
• E: Ecomimic Reform is Passed
• J: John is happy
• M: Mark is happy
• P: Paul is happy
Conver the following English sentences to Propositional Logic:
• If Smith wins the election and Economic Refrom is passed, then John will be happy
• If Smith loses the election and if Economic Reform is passed, then Mark will be happy
• Mark is not Happy
• Economic Reform is passed
• It is not true that Paul is happy if John or Mark is happy

Next, convert these sentences into Conjunctive Normal Form. Then use resolution to prove by refutation that John is happy. Show all your steps.