Anna University Chennai 2005 B.E Computer Science ARTIFICAL INTELLIGENCE & NEURAL NETWORKS - Question Paper

Code: C-20 / T-21 Subject: ARTIFICAL INTELLIGENCE &
NEURAL NETWORKS
Time: three Hours Max. Marks: 100
NOTE: There are 11 ques. in all.
ques. one is compulsory and carries 16 marks.
ans any 3 ques. every from Part I and Part II. every of these ques. carries 14 marks.
Any needed data not explicitly given, may be suitably presumed and said
Q.1 Answer the following: (2x8)

a. Describe briefly predicate logic knowledge representation scheme.

b. Assuming 'P' and 'R' to be false and 'Q' as true, evaluate and obtain out its truth value.

c. Discuss the 'frame problem' in the situation of dynamic environment.

d. Discuss how fuzzy matching can be accomplished ranging from objects.

e. Define a prolog function 'member (X, L)' which returns actual if 'X' is a member of list L, otherwise false.

f. Translate the subsequent text into predicate calculus formula :-
If 'A' is greater than 'B' and 'B' is greater than 'C' then 'A' is greater than 'C'.

g. Explain the difference ranging from 'knowledge' and 'belief'.
Q.4 a. Explain Hill climbing search strategy. List 3 issues which are encountered in this strategy. (8)

b. Using the search tree provided below, list the elements of the queue just before the next node is expanded. Use Best - 1st search where the numbers correspond to estimated cost to goal for every corresponding node. (6)


Q.5 a. Discuss unification and resolution in predicate logic using examples. (4)

b. Apply unification algorithm in every of the subsequent pair of literals and obtain most general unifier, if possible. (6)
(i) f (Marcus) and f (Caesar)
(ii) f (x) and f (g (y))
(iii) f (Marcus, g (x, y)) and f (x, g (Caesar, Marcus))

c. Under what conditions, best-first search would be worse than simple breadth-first search. (4)

Q.6 a. A issue solving search can proceed either using forward or backward chaining. What factors determine the option of direction for a particular problem? provide 2 examples - 1 using forward and other using backward chaining. (8)

b. Discuss the A* and AO* algorithms with suitable applications. (6)
PART II
ans any 3 ques.. every ques. carries 14 marks.
Q.7 a. How constraint satisfaction issues are various from state search problems? Compare them through example. (6)

b. Describe the subsequent knowledge representation techniques and compare them.

(i) Frames.
(ii) Semantic Nets. (8
Q.8 a. Give a simple model of a learning system. Illustrate through examples the subsequent learning techniques . (6)
(i) Decision tree induction. (ii) Learning by analogy.

b. Describe the back propagation learning algorithm pertaining to Neural Network. Compare it with Hopfield Networks. (8)

Q.9 a. Discuss the subsequent expert system architectures (10)
(i) Rule based systems. (ii) Non production systems.

b. What do you understand by knowledge acquisition? (4)

Q.10 a. Contrast expert system and Neural network in terms of knowledge representation, knowledge acquisition and explanation. (6)

b. Rule based systems often contain rules with several condition in the left sides. Why is this actual in MYCIN? (4)

c. Using MYCIN's rules for inexact reasoning, calculate CF, MB and MD of h, provided 3 observations where (4)

CF(h1,o1) = 0.5
CF(h1,o1) = 0.3
CD(h1,o3) = -0.2

Q.11 a. Discuss ATN & RTN based parsing. (6)

b. For every of the subsequent phrases. Show a parse tree and also discuss what knowledge in addition to the English grammar is necessary to produce the accurate parse. (8)

(i) Mohan wanted to go to the movie with Sohan.
(ii) Mohan wanted to go to the movie with Sohan Sandhya.
(iii) I heard the story listening to the radio.
(iv) I heard the children listening to the radio.


h. Describe briefly 1 Heuristic technique in AI.
PART I
ans any 3 ques.. every ques. carries 14 marks.
Q.2 a. Why is Prolog useful for developing simple expert system? (5)

b. Discuss briefly 3 applications of AI. (9)

Q.3 a. Enumerate at lowest 6 knowledge representation schemes. (6)

c. Write a Prolog function that returns a list found by appending a list 'A' to a different list 'B'. (8)

1
Index Next>>

Earning:   Approval pending.