Artificial Intelligence

Question 1

2014 4.a Give argument:

Answer 1

图翻译出formulae

Argument
Rebuts: 推翻claim
Undercuts：推翻support

Question 2

2014 4.b Definition of Diagnose over Abduction

Answer 2

Elements Set of facts(observations), set of background knowledge, set of defeasible rules, the diagnose is the subset T’ of T such that

1. 所有不能推出
2. 诊断可以推出
3. 诊断不冲突
4. 没有诊断子集可以推出

Question 3

2014 4.b 举例说明Diagnose:

Answer 3

Defeasible facts: d1, d2, d3, d4
Observation: o1
Problem: e1
Diagnosis: d1, d2
Into argument: 
* observation and problem is an atom
* observation is the start, problem is the end

Question 4

Breadth-frist search

Answer 4

Breadth-frist search: explore every node from the current node before going down a level.
Ad: Exhaustive
Disad: Require lots of memory, lots of work.

Question 5

Depth-first search

Answer 5

Depth-first search: going from the current node to the first available node on the next level. Continue until the can go no further. If goal state has not then been reached, then backtrack to last branch point.
Ad: Exhaustive
Disad: Expensive backtracking.

Question 6

Best-first search

Answer 6

Best-first: At each step, select the most promising of the nodes generated so far using heuristic functions. One of the mostly used is A* algorithm. g(n)=cost the the path from the current node to the initial node f(n)=estimated cost from the current node to the goal, usually given by straight line distance. Expand node with lowest f(n)=g(n)+h(n).
Guarantee of optimal: only if h never over-estimates.

Question 7

Write the lottery

Answer 7

[p, a; p, a]

Question 8

Give utility function based on different situation, expected Utility

Answer 8

consider carefully about the situation
U(情况) = 数值
正副不要对称，要有倾向性
EU(类)= sum of U