question bank-ai-12-13-10144cs601

30
10144CS601 ARTIFICIAL INTELLIGENCE QUESTION BANK 1 UNIT I Short Questions 1) Define AI Artificial Intelligence is the study of how to make computers do things which, at the moment, people do better. 2) What are the requirements by a computer to pass a Turing test? The requirements by a computer to pass a Turing test are : Natural language Processing , Knowledge representation Automated reasoning , Machine learning 3) When is a class of problems said to be intractable ? A class of problems is said to be intractable if the time required to solve instances of the class grows at least exponentially within the size of the instances. 4) List some of the AI tasks Some of the AI tasks are, Vision, Natural language, Planning, Robotics, Medical diagnostics Equipment repair, Computer configuration, Financial planning 5) What are the main tasks in entire language processing problem? Main tasks in entire language processing problems are: Processing written text, Processing spoken language 6) Define an agent. An agent is anything that can be viewed as perceiving its environment through sensors and acting upon that environment through effectors. (OR) It is a physical object that can be analysed as having perception and producing actions. 7) Define rational agent A rational agent always selects an action based on the percept sequence it has received so as to maximize its (expected) performance measure given the percepts it has received and the knowledge possessed by it. 8) Distinguish between deterministic and stochastic environment. Deterministic Stochastic The next state of the environment is completely determined by the agent The environment is complex type with respect to the point of view of the agent The agent need not worry about the possible uncertainty. Since the environment is stochastic, the agent makes stochastic actions. 9) Give an informal description of the general search algorithm.

Upload: bhaskar-veeraraghavan

Post on 21-Jan-2016

178 views

Category:

Documents


0 download

DESCRIPTION

This is the question bank for the subject Artificial Intelligence 10144CS601 prescribed by AU 6th semester

TRANSCRIPT

Page 1: question bank-AI-12-13-10144CS601

10144CS601 ARTIFICIAL INTELLIGENCE

QUESTION BANK

1

UNIT – I

Short Questions

1) Define AI

Artificial Intelligence is the study of how to make computers do things

which, at the moment, people do better. 2) What are the requirements by a computer to pass a Turing test?

The requirements by a computer to pass a Turing test are :

Natural language Processing , Knowledge representation

Automated reasoning , Machine learning 3) When is a class of problems said to be intractable ?

A class of problems is said to be intractable if the time required to solve

instances of the class grows at least exponentially within the size of the

instances. 4) List some of the AI tasks

Some of the AI tasks are,

Vision, Natural language, Planning, Robotics, Medical diagnostics

Equipment repair, Computer configuration, Financial planning

5) What are the main tasks in entire language processing problem?

Main tasks in entire language processing problems are:

Processing written text, Processing spoken language

6) Define an agent.

An agent is anything that can be viewed as perceiving its environment

through sensors and acting upon that environment through effectors. (OR)

It is a physical object that can be analysed as having perception and

producing actions.

7) Define rational agent

A rational agent always selects an action based on the percept sequence it has

received so as to maximize its (expected) performance measure given

the percepts it has received and the knowledge possessed by it.

8) Distinguish between deterministic and stochastic environment.

Deterministic Stochastic

The next state of the environment is

completely determined by the agent

The environment is complex type

with respect to the point of view of

the agent

The agent need not worry about the

possible uncertainty.

Since the environment is stochastic,

the agent makes stochastic actions.

9) Give an informal description of the general search algorithm.

Page 2: question bank-AI-12-13-10144CS601

10144CS601 ARTIFICIAL INTELLIGENCE

QUESTION BANK

2

function GENERAL-SEARCH(problem, strategy)returns a solution or

failure

initialize the search tree using the initial state of problem

loop do

if there are no candidates for expansion then return failure

choose a frontier node for expansion according to strategy if the node contains a goal state then return the corresponding solution

else expand the node and add the resulting nodes to the search tree

end

10) What are the factors to measure the performance of searching techniques? Completeness, time complexity, space complexity and optimality

11) What is the application of best first search?

Uniform Cost Search is a special case of the A* algorithm. The algorithm

maintains a priority queue of nodes to be explored. A cost

function f(n) is applied to each node. The nodes are put in OPEN in the order

of their f values. Nodes with smaller f(n) values are expanded earlier.

12) Name the types of search strategies. There are two types of search strategies. They are,

a. Uninformed search or blind search 1. Breadth-first search

2. Uniform cost search

3. Depth-first search

4. Depth-limited search

5. Iterative deepening search

6. Bi-directional search

b. Informed search or heuristic search 13) Give the advantages of BFS (Breadth First Search) BFS was guaranteed to find the shortest solution path but required

inordinate amounts of space because all leaf nodes had to be kept

in memory. 14) Give the advantages of DFS (Depth First Search) DFS was efficient in terms of space but required some cut-off depth in

order to force backtracking when a solution was not found.

15) Define production system. A set of condition action rule is termed as a production system which

incorporates the forward chaining. 16) What elements are necessary to formally define a search problem? Initial state, goal test and operator description.

Page 3: question bank-AI-12-13-10144CS601

10144CS601 ARTIFICIAL INTELLIGENCE

QUESTION BANK

3

17) Develop a PEAS description of the task environment for an internet

book- shopping agent or an interactive English tutor Page 40 AIMA 2/e(your text book)

Performance measure:-Get the required book within the least time and at

the least cost and within the budget.

Environment:- Internet

Actuators :- Ability to locate the desired link for the book.

Sensor:-Ability to understand if the desired book is available. 18) State True/False & justify your answer:

(a)There exists task environments (PEAS) in which some pure reflex

agents behave rationally.

(b) The input to an agent program is the same as the input to the

corresponding agent function.

(c) There exists task environments (PEAS) in which some pure reflex

agents behave irrationally. (a) True: Page 48 AIMA 2/e(your text book)

(b) False: the agent program takes the current percept whereas the agent

function takes the percept history. (c) True. 19) What do you mean by or what is meant by belief state? (Page 84) in your text 20) Describe a state space in which iterative deepening search performs much

worse than depth-first search(for example O(n2) vs. O(n)).

Consider a domain in which every state has a single successor, and there

is a single goal at depth n. Then the depth-first search will find the goal in

n steps, whereas iterative deepening search will take 1+2+3+…+ n =

O(n2) steps

21) State True/False and justify: The agent that senses only partial

information about the state cannot be rational. False: the rationality and omniscience are different concepts. 22) Explain why problem formulation must follow goal formulation? 23) What is a complete search algorithm?

An algorithm which guarantees to find the goal(or solution) if one exists

at a finite depth. 24) Why does one go for heuristic search? Heuristic means “rule of thumb”. To quote, Judea Pearl, “Heuristics are

criteria, methods or principles for deciding which among several

Page 4: question bank-AI-12-13-10144CS601

10144CS601 ARTIFICIAL INTELLIGENCE

QUESTION BANK

4

alternative courses of action promises to be the most effective in order to

achieve some goal”. In heuristic search or informed search, heuristics are

used to identify the most promising search path

25) Describe a knowledge-based agent at three levels.(sorry, Unit 2 Question) a. Knowledge level or epistemological level, describe the agent by

saying what it knows

b. Logical level, at which the knowledge is encoded into sentences.

c. Implementation level, that runs on the agent architecture. 26) What is a heuristic? A heuristic is a technique that improves the efficiency of the search

process, possibly by sacrificing claims of completeness. In other words, it

is an estimate which may guide to reach the goal with an optimism,

without much prejudice to its possible completeness. 27) What is a heuristic function ? A heuristic function is a function that maps from the problem state

description to measures of desirability, usually represented in numbers. 28) What is the purpose of the heuristic function? To guide the search process in the most profitable direction by suggesting

which path to follow first when more than one is available.

29) Define consistency or monotonicity of a heuristic. A heuristic h(n) is consistent if, for every node n and every successor n, of

n generated by any action a, the estimated cost of reaching the goal from n

is no greater thanthe step cost of getting to n plus the estimated cost of

reaching the goal from n: h(n) c(n,a,n) + h(n) 30) What is the condition for optimality of A* search algorithm? The heurist defined for the A* must be admissible and consistent. 31) How can we measure the effectiveness of a search? (i) Branching factor (ii) Depth (iii) Completeness 32) How can we choose the ideal search? We have seen several search strategies – BFS, DFS, Best first search and

so on – that are designed by computer scientists. There is a method rests

on an important concept called metalevel state space. By means of

meta-level learning algorithms, we can avoid exploring unpromising sub-

trees.

33) Compare A* search with IDA* in terms of space and time complexity.

Assuming a search tree with a uniform branching factor of b.

Page 5: question bank-AI-12-13-10144CS601

10144CS601 ARTIFICIAL INTELLIGENCE

QUESTION BANK

5

Algorithm/complexity TIME SPACE

A* bn

bn

IDA* b*bn

bn

34) List the advantages of informed search strategies. i) Informed search finds the goal node more efficiently than the

uninformed strategy.

ii) Using the evaluation function, the best first search for example finds

the lowest cost path to the goal node at the shortest time. The

local search algorithms could find the path by modifying the one or

more current states rather than doing a systematic search.

iii) The online search finds the goal node in a completely unknown

environment. 35) Name some local search algorithms. (i) Hill Climbing (ii) Simulated annealing (iii) Local beam search (iv)

genetic algorithm.

36) What is meant by plateau with respect to search techniques? A plateaux is an area of the state space where the evaluation function is

flat. That is all the neighbour nodes evaluates to the same value. 37) How to overcome the drawback of climbing local maxima/minima and

get stuck in a hill-climbing search to reach the global maxima/minima? Do a random restart hill climbing.

38) Mention the three drawbacks of hill climbing search. (i) local maxima (ii) Plateaus (iii) Diagonal ridges 39) Is hill climbing algorithm guaranteed to find a solution for the 8-queens

problem? No it can not find an optimum solution for every value of the initial state. 40) Is simulated annealing algorithm guaranteed to find a solution for the 8-

queens problem or any TSP problem? Even simulated annealing can get stuck at a local minima. Even otherwise

it is not guaranteed that it will find an optimal solution. 41) Tell about the principle behind simulated annealing.

First it does a random walk choosing neighbours at random and deciding

at random whether to visit that neighbour etc..

42) Then what is beam search? In hill climbing and stochastic algorithm only single node is considered

to be used but in Beam search where up to k nodes are maintained.

Stochastic Beam search is another variation of Beam search.

Page 6: question bank-AI-12-13-10144CS601

10144CS601 ARTIFICIAL INTELLIGENCE

QUESTION BANK

6

43) What about genetic algorithm?

First create a population then select parents, cross-over then produce a

child check the fitness and goal-test. if goal return; If not goal and fits ok,

do mutation, then add to new population and repeat the above process. 44) What is LRTA* search? See book. 45) Define a constraint satisfaction problem Constraint satisfaction problem or CSP is defined by a set of variables,

X1, X2, …. Xn, and a set of constraints, C1, C2,… Cm. Each variable Xi has

a non-empty domain Di of possible values. Each constraint Ci involves

some subset of variables and specifies the allowable combinations of

values for that subset. A state of the problem is defined by the

assignment of values to the some or all of the variables. The assignment

that does not violate any constraint is called a consistent or legal

assignment. A complete assignment is one in which every variable is

mentioned, and a solution to a CSP is a complete assignment that

satisfies all the constraints. 46) What is a CSP? A constraint satisfaction problem is a problem in which the goal is to

choose a value for each of a set of variables, in such a way that the values

all obey a set of constraints. A constraint is the restriction on the possible

values of two or more variables. 47) What is a Boolean CSP? Boolean CSPs are known as finite-domain CSPs, whose variables can be

either true or false.

48) What is forward checking? Whenever a variable X is assigned, the forward checking process looks at

each unassigned variable Y that is connected to X by a constraint and

deletes from Y’s domain any value that is inconsistent with the value

chosen for 49) What is backtracking search? It is a form of DFS in which there is a single representation of the state

that gets updated for each successor, and then must be restored when a

dead end is reached. 50) What are the two methods of performing CSP on a tree? (i) Variable or value ordering (ii) Forward checking(constraint

propagation)

Page 7: question bank-AI-12-13-10144CS601

10144CS601 ARTIFICIAL INTELLIGENCE

QUESTION BANK

7

51) Explain why it is a good heuristic to choose the variable that is most

constrained and the value that is least constraining in a CSP search? Because MCV chooses the variable that is likely to cause a failure, and is

most efficient to fail as early as possible thereby pruning large parts of

the search space. The least contraining value heuristic is better because it

allows the most chances for the future assignments to avoid conflicts.

Part B Questions

1) Explain the structure of agents with neat diagram 4 2) Prove that the vacuum agent is rational 6

It suffices to show that for all possible actual environments (i.e., all dirt

distributions and initial locations), this agent cleans the squares at least as

fast as any other agent. This is trivially true when there is no dirt. When

there is dirt in the initial location and none in the other location, the

world is clean after one step; no agent can do better. When there is no dirt

in the initial location but dirt in the other, the world is clean after two

steps; no agent can do better. When there is dirt in both locations, the

world is clean after three steps; no agent can do better. (Note: cleaning

squares at least as fast as any other agent is a much stricter condition than

necessary for an agent to be rational; it just needs to be as good as any

other agent with the same sensors and actuators.) 3) Give the initial state, goal test, successor function, and cost

function for each of the following. Choose a formulation that is

precise enough to be implemented.

(a) You have to colour a planar map using only three colours,

with no two adjacent regions having the same colour.

4

This is a version of the graph-colouring problem (a standard CSP)

described in lecture.

Initial state: An uncoloured map.

Goal test: Each region on the map is shaded in one of (at most) three

colours, and no region shares a colour with its adjacent regions.

Successor function: For a partially-coloured map, the successor of a

particular partially-coloured map, on which an agent colours a new region,

is a copy of the map with the new region coloured in.

Cost function: If the cost of colouring a region is independent of the colour

used, all paths to the goal will have the same cost. No solution is preferred

Page 8: question bank-AI-12-13-10144CS601

10144CS601 ARTIFICIAL INTELLIGENCE

QUESTION BANK

8

over any other, so let the cost of colouring a state be 1.

4) Discuss about searching with partial information. 8 5) Explain A* search algorithm in detail. 8 6) What are AND-OR graphs? Explain. 8 7) Explain AO* search algorithm in detail(refer to Principles of AI

by Nils Nillsson which is at library Self-study Topic)

8

8) Discuss the problems(drawbacks) in hill-climbing algorithm.

What is local beam search?

8

9) Explain Genetic Algorithm in detail. What are the operators of

genetic programming?

8

10) Devise a state space in which A* using GRAPH-SEARCH

returns a suboptimal solution with an h(n) function that is

admissible but inconsistent.

4

In the above graph, GRAPH-SEARCH expands A first rather

than B and returns a suboptimal solution. h(n) is an inconsistent

heuristics.

11) Solve the crypt arithmetic problem given in page 141(fig. 5.2).

by hand using backtracking, forward checking.

8

12) Discuss the various issues associated with the backtracking 8

Page 9: question bank-AI-12-13-10144CS601

10144CS601 ARTIFICIAL INTELLIGENCE

QUESTION BANK

9

search for CSPs

13) Give a precise formulation for each of the following CSP

problems:

i) Rectilinear floor planning: find nonoverlapping places in a

large rectangle for a number of smaller rectangles.

ii) Class scheduling: There are a fixed number of professors and

classrooms, a list of classes to be offered, and a list of possible

time slots for classes. Each professor has a set of classes that

he/she can teach.

8

14) Solve the 8-Puzzle using A* algorithm and find the heuristic

function

UNIT II

Short Questions

1) Define logic in general. Logics are formal languages for representing information such that

conclusions can be drawn. 2) Define syntax and semantics in logic. Give examples.

Syntax defines the acceptable sentences in the language (called wffs or

well formed formulas) and Semantics define the "meaning" of sentences

i.e., how to decide truth of a sentence in a world model.

E.g., in the language of arithmetic

x+2 > y is a sentence; x2+y > is not a sentence (syntax)

x+2 > y is true iff the number x+2 is greater than the number y

x+2 > y is true in a world where x = 7, y = 1

x+2 > y is false in a world where x = 1, y = 7 3) Define entailment. How to represent it? How to prove it? We want to generate new sentences that are necessarily true, given that

the old sentences are true. This relation between sentences is called

entailment.

In other words, entailment means that one thing follows from another:

KB ╞ α

Knowledge base KB entails sentence α if and only if α is true in all

world models where KB is true

E.g., the KB containing “the women are unsafe in Delhi” and “the men

Page 10: question bank-AI-12-13-10144CS601

10144CS601 ARTIFICIAL INTELLIGENCE

QUESTION BANK

10

are safe in Chennai” entails “The women are unsafe in Delhi and the

men are safe in Chennai ”

The entailment can be proved by (i) Truth Table method and (ii)

Inference method.

4) Give a short note on inference. Let KB ├i α = sentence α can be derived from KB by an algorithmic

procedure i. It is a relationship between sentences based on their

syntax.

Soundness: i is sound if whenever KB ├i α, it is also true that KB╞ α

Completeness: i is complete if whenever KB╞ α, it is also true that

KB ├i α which means that is derived from KB using i. Here i is known

as the inference algorithm. 5) Give a pictorial representation of proof theory and model theory of

logic.

6) What are the functionalities of a knowledge-based agent?

It has 3 aspects to perform:

a. Background knowledge – TELL & ASK, MAKE-PERCEPT-SENTENCE

& MAKE-ACTION QUERY

b. Knowledge level – what(or which) will achieve its goal etc.

c. Implementation level – declarative and procedural approach.

7) Name the two kinds of agents that are built based on Propositional logic. (i) Ones that uses inference algorithms and a KB.

(ii) The ones which evaluate logical expressions directly in the form

of circuits.

Page 11: question bank-AI-12-13-10144CS601

10144CS601 ARTIFICIAL INTELLIGENCE

QUESTION BANK

11

8) Explain Modus Ponen and And –Elimination

They are inference rules and is written as follows :

,

the notation means that, whenever any sentence of the form and

are given, then the sentence can be inferred. For example, if

(WumpusAhead WumpusAlive) shoot and

(WumpusAhead WumpusAlive) are given, then shoot can be inferred.

Another inference rule is And-Elimination, which says that, from

a conjunction, any of the conjuncts can be inferred:

for example, from WumpusAhead WumpusAlive), WumpusAlive

can be inferred.

9) Write the seven inference rules for propositional logic. The seven inference rules for propositional logic are,

1. Modus ponens

2. And-elimination

3. And-introduction

4. Or-introduction

5. Double-negation elimination

6. Unit resolution

7. Resolution 10) Define the interpretation in propositional logic. Given a propositional formula G, let

{A1,...,An} be the set of atoms which occur in the formula, an

Interpretation I of G is an assignment of truth values to {A1,...,An}. 11) When do we say an interpretation I a model of the formula G? If G is true under the interpretation I then I is a model for G and we can

write I ⊨ G 12) Define validity and satisfiability.

A formula F is valid if an only if it is true in all its interpretation. A

formula F is inconsistent if an only if it is false in all its interpretation.

A formula which is not inconsistent is consistent or satisfiable. 13) What is meant by decidability? If there exists a terminating method to decide whether a formula is valid. 14) Define what is a literal.

Page 12: question bank-AI-12-13-10144CS601

10144CS601 ARTIFICIAL INTELLIGENCE

QUESTION BANK

12

A literal is an atom or the negation of an atom.

15) Define CNF. A formula F is in Conjunctive Normal Form (CNF) iff it is in Negation

Normal Form and it has the form F = F1 F2 ... Fn, where each Fi is a

disjunction of literals. Here each Fi is called the clause. 16) What is a clause? A clause is a disjunction of literals. 17) Define predicate logic. Predicate logic is defined as,

1. Predicate logic allows us to represent fairly complex facts about the

world, and to derive new facts in a way that guarantees that, if the initial

facts were true, then so are the conclusions.

2. There are two quantifications used in addition to propositional logic

the are

there exists() and for all () 18) What are the two quantifiers in predicate logic? Give example. The two quantifiers in predicate logic are and , the following are valid

sentences:

x. Bird(x) Dontfly(x) – There exists a bird that doesn‟t fly.

x. Bird(x) hasfeather(x) – Every bird has feather.

19) Define an atom in FOL.

i. Individual symbols or constants: they are usually the names of objects

such as John, Mary and 3.

ii. Variable symbols or constants: they are either lowercase or

uppercase(in Prolog) x,y,z, ... or (X, Y, Z, ...)

iii. Function symbols: these are customarily lowercase f, g, h ...

iv. Predicate symbols: These are customarilyuppercase but lowercase as

in Prolog. P, Q, R, GREATER and LOVE. 20) Distinguish bet ween free and bound in a formula.(scope) An occurrence of a variable in a formula is bound if and only if the

occurrence is within the scope of a quantifier employing the variable, or

is the occurrence in that quantifier. An occurrence of a variable in a

formula is free if and only if the occurrence of the variable is not bound.

21) Give a formal definition of a formula in FOL. i. An atom is a formula(note that the atom is an abbreviation of

atomic formula)

Page 13: question bank-AI-12-13-10144CS601

10144CS601 ARTIFICIAL INTELLIGENCE

QUESTION BANK

13

ii. If F and G are formulas, then (F), (F G), (F G ),( F G)and

(F G) are formulas.

iii. If F is a formula, and x is a variablein F, then (x) F and (x) F

are formulas.

iv. Formulas are generated by finite number of applications of (i), (ii)

and (iii) above.

22) Symbolise the following sentences:

(a) Every rational number isa real number.

(b) There exists a number that is a prime. Ans:

(a) ( x) (Q(x) R(x))

(b) ( x) P(x) 22) Define – sound inference rule. A pattern of inference that is guaranteed to generate true conclusions

given true premises 23) Which of the following is entailed by the sentence

(A B) (C D E)

(a)(A B) (b) ( A B C) ( B C D E)

(c)(A B) (D E) Ans:

(a) ENATILED. Simple AND – elimination.

(b) ( B C D E) is equivalent to (B C D E) so

this simply weakens the clause by introducing another disjunct.

(c) (A B) (D E) - NOT ENTAILED - this removes the

C literal, which strengthens the clause.

24) What is meant by logical consequence?

Given a set of formulas {F1,... Fn} and a formula G, G is said to be a logical

consequence of F1,..., Fn iff for any interpretation I in which

F1 F2 ... Fn is true G is also true.

If G is a logical consequence of {F1,... Fn} we write F1 F2 ... Fn ⊨ G

F1,..., Fn are called the axioms or premises for G.

25) Define unification.

The unification routine, UNIFY is to take two atomic sentences p and q and

return a substitution that would make p and q look the same.

UNIFY (p,q) =

Where SUBST ( ,p) = SUBST (,q)

Page 14: question bank-AI-12-13-10144CS601

10144CS601 ARTIFICIAL INTELLIGENCE

QUESTION BANK

14

26) Define resolution.

The most well known general proof procedure for predicate calculus

is resolution. Resolution is a sound proof procedure for proving things

by refutation-if you can derive a contradiction from p then p must be true.

27) Define Herbrand’s universe.

If S is a set of clauses, then HS the Herbrand’s universe of S, is the set of all ground terms constructible from the following: a. The function symbols in S, if any.

b. The constant symbols in S, if any; if none then the constant symbol A.

e.g.: if S contains just the clause P(x, F(x, A)) Q(x,A) R(x, B), then

HS is the following infinite set of ground terms:

{A, S, F(A,A), F(A,B), F(B,A), F(B,B), F(A,F(A,A)),…}. 28) What are skolem function? Functions with the same number of arguments must be generated as

the number of universal quantifiers in whose scope the expression occurs.

These generated function are called skolem functions.

29) When do you say the two objects match? In higher order logic, two objects are equal if and only if all properties

applied to them are equivalent.

x,y (x = y) ( P P(x) P(y)) 30) What is lifting lemma?

Let C1 and C2 be two clauses with no shared variable, and let C1 and C2 be

ground instances of C1 and C2. If C is a resolvent of C1, then there exists a

clause C such that (1) C is a resolvent of C1 and C2 and (2) C is a ground instance of C. This is called lifting lemma because, it lifts the proof steps from ground clauses up to general first-order clauses.

31) What is subsumption?

If P(A) Q(B) is in the KB, then it is not necessary to add P(A) to the KB

because, P(A) subsumes P(A) Q(B). This feature is called

subsumption. 32) State True/False:

a. x,y. x = y is satisfiable.

b. (x P (x)) (x P (x)) is a valid sentence. Every valid sentence

is satisfiable.

c. The only unsatisfiable clause is the empty clause.

Page 15: question bank-AI-12-13-10144CS601

10144CS601 ARTIFICIAL INTELLIGENCE

QUESTION BANK

15

d. If C1 and C2 are clauses in propositional logic and all the literals in

C1 are contained in C2, then C1 ╞═ C2. e. If C1 and C2 are clauses in propositional logic C1 ╞═ C2, then all the

literals in C1 are contained in C2

f. The clause A B C entails the clause (A B).

g. Forward chaining is a complete inference procedure. Ans: (a) TRUE: satisfied by a model with exactly one object. (b) False

(c) True (d) True (e) True (f) True (g) False; the entailment is the other

way.(h)false. It does not have resolution proof.

Part B Questions

1) Describe the Wumpus world environment in detail. 8

2) Explain the forward chaining algorithm for the propositional

logic.(AUTT M/J 2009)

8

3) Explain the backward chaining algorithm for the PL.(M/J 2009) 8

4) What is Davis Putnam Algorithm? Describe DPLL for PL. 8

5) Explain how to find pits and Wumpuses using logical

inference?

8

6) Consider the following facts: (AUTT M/J 2009)

John likes all kinds of food.

Apples are food.

Chicken is food

Anything anyone eats and isn‟t killed is food.

Bill eats peanuts and is still alive.

Sue eats everything Bill eats.

(i) Translate these sentences in formulas in predicate logic.

(ii) Prove that john likes peanuts using backward chaining.

(iii) Convert the formula of a part in clausal form.

(iv) Prove that john likes peanuts using resolution.

16

Ans (ii)

Page 16: question bank-AI-12-13-10144CS601

10144CS601 ARTIFICIAL INTELLIGENCE

QUESTION BANK

16

7) Using the facts:

(i) Rama was a man

(ii) Rama was a saint

(iii) All saints are Hindus

(iv) Ravana was ruler

(v) All Hindus were either loyal to Ravana or hated him.

(vi) Everyone is loyal to someone

(vii) People only try to assassinate rulers they are not loyal to.

10

Page 17: question bank-AI-12-13-10144CS601

10144CS601 ARTIFICIAL INTELLIGENCE

QUESTION BANK

17

(viii) Rama tried to assassinate Ravana

(ix) All men are people.

Answer the question: “Did Rama hate Ravana?” using

Resolution (AUTT M/J 2011)

8) Explain the steps in knowledge Engineering process in FOL. 8

9) Unify the following pairs(if possible)

a. P(A, B, B) , P(x, y, z)

b. Q(y, G(A,B)), Q(G(x,x),Y).

c. Older(father(y), y), Older(father(x) John).

6

Answer

a. {x/A,y/B, z/B}

10) Discuss on the “completeness of resolution”. 6 11) Translate into FOL the following sentences:

i. “If someone understands someone, then he is that someone’s

friend”

ii. “Friendship is transitive”

iii.“All German speak same languages”

i. x,y Understyands(x,y) Friend(x,y)

ii. x,y,z. Friend(x,y) Friend(y,z) Friend(x,z)

iii. x, y, l German(x) German(y) Speaks(x, l) Speaks(y ,l). 12) 13) 14)

Page 18: question bank-AI-12-13-10144CS601

10144CS601 ARTIFICIAL INTELLIGENCE

QUESTION BANK

18

UNIT III

Short Questions and answers (more questions may be given later)

1) Describe the differences and similarities between problem solving and

planning In planning we have the goal state decomposable into several sub-goals.

Again the states are decomposable. An action changes only a few of the

variables.

2) Define planning with state space search

In planning algorithm we have several actions each associated with set

of preconditions and effects. We thus can choose appropriate actions

from the start state to the final state in a forward search or progression

and from the final to initial in a backward search or regression.

3) What is an action schema?

Representation of actions. An action is specified in terms of the

preconditions that must hold before it can be executed and the effects that

ensue when it is executed. For example, an action for flying a plane from

one location to another is:

Action(Fly(p; from; to);

PRECOND:At (p; from) Plane(p) Airport(from) Airport(to)

EFFECT::At (p; from) At (p; to))

This is more properly called an action schema, meaning that it represents a

number of different actions that can be derived by instantiating the variables

p, from, and to to different constants. In general, an action schema consists

of three parts 1. Action name and parameter list. 2. The precondition.

3. The effect. 4) What is a planning graph?

A planning graph consists of sequence of levels that correspond to time

steps in the plan, where level 0 is the initial state. Each level contains a

set of literals and a set of actions that could be true at that time step

depending on the action taken in previous time steps. For every positive

and negative literals c we add a persistence action with precondition C

and effect C. 5) What is progression planning?

Search queue:{Initial State}

Loop

Pick a state from the search queue

If it is a goal state, terminate, and return the path from the init. State.

Page 19: question bank-AI-12-13-10144CS601

10144CS601 ARTIFICIAL INTELLIGENCE

QUESTION BANK

19

Apply all the applicable actions in the state

“Progress the state through the operator application

Put each of the states in the search queue

End. 6) Mention the problems with forward state space search and how to

overcome them? Branching factor is the number of legal actions. Path length is the

number of actions required to achieve the goal. In real world

applications this will amount to too many applicable actions. To

overcome the above, we must apply heuristics or search in backward

direction.

7) What is regression planning It is nothing but backward search. Given a goal, we consider only

actions that actually achieve it. 8) Mention the advantage of backward search while planning

9) Mention the difference between STRIPS and ADL

STRIPS Action Description Language

Only ground literals in goals

Fat Slow

Quantified variables in goals

Closed World Assumption:

Unmentioned literals are false

Open World Assumption:

Unmentioned literals are

unknown

10) Name the advantage of ADL over STRIPS Pg. 378-79(expressiveness of Planners)

11) Heuristics for state space search By finding the number of false goals propositions attained by the

current state. This can be found only for relaxed problems. 12) What is sub goal independence assumption

The cost of solving a conjunction of sub-goals is approximated by the

sum of the costs of each goal independently. 13) What is empty – delete – list heuristic

It is also possible to generate relaxed problems by removing negative effects

without removing preconditions. That is, if an action has the effect A B

in the original problem, it will have the effect A in the relaxed problem. This

means that we need not worry about negative interactions between subplans,

Page 20: question bank-AI-12-13-10144CS601

10144CS601 ARTIFICIAL INTELLIGENCE

QUESTION BANK

20

because no action can delete the literals achieved by another action. The

solution cost of the resulting relaxed problem gives what is called the

empty-delete-list heuristic.

14) What is least commitment strategy The general strategy of delaying a choice during search is called a least

commitment strategy. There is no formal definition of least commitment,

and clearly some degree of commitment is necessary, lest the search would

make no progress. 15) Define partial order planner (June 07)

If a planning algorithm can place any of the actions into the plan without

specifying which comes first is called a partial order planner.

16) What are planning graphs? All of the heuristics we have seen for total-order and partial-order planning

can suffer from inaccuracies. A special data structure called a planning

graph can be used to give better heuristic estimates. 17) State the condition for the termination of GRAPHPLAN

Literals increase monotonically

Actions increase monotonically

Mutexes decrease monotonically

If all the above stated properties are noticed, GRAPHPLAN will terminate.

18) What is a SATPLAN algorithm?

In SATPLAN, we will be finding models of propositional sentences that

look like this:

initial state all possible action descriptions goal

The sentence will contain proposition symbols corresponding to every

possible action occurrence; a model that satisfies the sentence will assign

true to the actions that are part of a correct plan and false to the others. An

assignment that corresponds to an incorrect plan will not be a model,

because it will be inconsistent with the assertion that the goal is true. If the

planning problem is unsolvable, then the sentence will be unsatisfiable. 19) What is planning and acting in real world

20) What is Critical Path Method (CPM)

Given a partial ordering of actions with durations , we can apply the

critical path method to determine the possible start and end times of

each actions. The critical path is the path whose total duration is the

longest.

Page 21: question bank-AI-12-13-10144CS601

10144CS601 ARTIFICIAL INTELLIGENCE

QUESTION BANK

21

21) What is HTN planning?

Here plans are generated by only successive action decompositions. 22) Name the different methods of planning and acting in nondeterministic

domains. Sensorless planning or conformant planning, conditional planning or

contingency planning, execution monitoring and re-planning and

continuous planning

PART B Questions

1) Explain the concept of Planning with state space search with

suitable examples.

8

2) What is STRIPS explain in detail with the example 8 3) Explain partial order planning with example 8 4) Explain the use of planning graphs in providing better heuristic

estimates with examples

8

5) How we plan and act in non deterministic domains 8 6) Write short notes on the following: 16

i) Conditional planning

ii) Execution monitoring and replanning

iii) Continuous planning

iv) Multiagent planning

7) Give an example for partial order planning 8 8) what is Backward state space search 8 9) Explain Heuristics for state space search 8

10) Explain Forward state space search 8 11) For Air cargo transport explain STRIPS 8 12) For Blocks World explain STRIPS 8 13) Compare STRIPS and ADL language 8

14) How we plan with propositional logic 8 15) Explain any two real world applications of planning and acting 8 16) How we schedule with resource constraints 8 17) Explain partial order planning with unbound variables 8 18) What are planning graphs? Explain the method of planning

and acting in real world.

16

Page 22: question bank-AI-12-13-10144CS601

10144CS601 ARTIFICIAL INTELLIGENCE

QUESTION BANK

22

UNIT IV

Short Questions & answers

1) Define uncertainty (june 07)

When the agent does not have an absolute knowledge on the outcome

of its actions, we say that the world or environment is uncertain or

stochastic. In such circumstances it is necessary that the agent applies

the probabilistic logic to deal with such uncertainty.

2) Mention the difference between statistics and probability.

Statistics is learning and probability is inference using evidence. 3) Define Baye’s rule (june 06)

Using product rule,

P(A B) = P(A | B)P(B) = P(B | A)P(A) Bayes rule

P(A | B) =

P(B|A) is called the likelihood; P(A) is called the prior and P(B) is

called the marginal likelihood. P(A|B) is the posterior.

P(B) = which is the total probability or

graphically,

The term P(B) is hard to compute.

A

B

A is Not observable

A causes B to occur

B is Observable

Diagnostic Reasoning

Page 23: question bank-AI-12-13-10144CS601

10144CS601 ARTIFICIAL INTELLIGENCE

QUESTION BANK

23

4) How is uncertainty knowledge represented ? Give an example.

Consider the rule,

A B is interpreted as the conditional probability

expression P(B | A) = m, stating that among all worlds satisfying A,

those that also satisfy B constitute a fraction of size m. In Dempster –

Shafer formalism, the above rule does not describe how an agent reacts

to the finding of A, but asserts that the set of worlds in which A and B

hold simultaneously has low likelihood and hence should be excluded

with probability m. 5) What are the axioms of probability

All probabilities lie between 0 and 1

0 P(a) 1

P(true) = 1 and P(false) = 0

P(a b) = P(a) + P(b) – P(a b)

The above three are also known as Kolmogorov’s axioms 6) Define Decision Theory

Decision theory = probability theory + utility theory. Also, an agent is

rational if it chooses the action that yields the highest expected utility,

averaged over all the possible outcomes of the actions. 7) Define probabilistic inference

It is nothing but the semantics of Belief or Bayes networks. 8) What is Markov blanket

The Markov blanket of a node is its set of neighboring nodes: its

parents, its children, and any other parents of its children. X is a

Bayesian network with respect to G if every node is conditionally

independent of all other nodes in the network, given its Markov

blanket. 9) What is noisy logical relationship.

This is used while modelling multicausal interactions. If Ui is the only

parent that is TRUE, X will be TRUE iff the inhibitor associated with Ui

remains inactive. Thus we have

P(X = TRUE | Ui = TRUE, Uk = FALSE k i) = 1 - qi .

The parameter

Ci = 1 - qi

represents the degree to which a isolated explanation Ui = TRUE can

endorse the consequent event. This is known as a noisy logical

m

m m

Page 24: question bank-AI-12-13-10144CS601

10144CS601 ARTIFICIAL INTELLIGENCE

QUESTION BANK

24

relationship. This method is very efficient in medical world.

10) What is a Temporal Model

Using Et = set of observable evidence variables at time t

If we can model

Xt = set of unobservable state variables at time t for each time step,

such a model is known as temporal model. 11) Define HMM

Markov assumption: Xt depends on bounded subset of X0:t – 1

each state depend on the previous state only.

First order Markov process is:

P(Xt | X0:t – 1 ) = P(Xt | Xt – 1)

12) What is smoothing

It is a technique used in recognising speech signals using the stationary

first order Markov chains or bigrams and second order Markov chains

or trigrams for different English words. 13) What is hindsight 14) Define EM algorithm

EM algorithm is a statistical learning method for recognising speech

etc.. 15) How to handle uncertain knowledge

Refer text book

16) What are the basic probability notation 17) What is prior probability 18) Distinguish between full joint probability distribution and joint

Probability distribution 19) Write an algorithm for decision theoretic agent

Refer text book 20) 21)

PART B Questions

1) What is Baye’s rule ? explain how Baye’s rule can be applied

to tackle uncertain knowledge (june 07)

2) Explain probabilistic reasoning ( june 07) 3) Explain HMM 4) What is a Bayesian network

Page 25: question bank-AI-12-13-10144CS601

10144CS601 ARTIFICIAL INTELLIGENCE

QUESTION BANK

25

5) How to get the exact inference form Bayesian network

6) How to get the approximate inference form Bayesian network 7) Explain all the temporal models 8) How to act under uncertainty 9) In temporal model explain filtering and prediction

10) Explain Smoothing with needed algorithm 11) How to handle uncertainty 12) How to construct Bayesian network 13) What are all the exact inference in Dynamic Bayesian Network 14) What are all the approximate inference in DBN 15) How to represent knowledge in an uncertain domain

16) Explain the use of Hidden Markov Models in Speech

Recognition in detail

17) Explain in detail about (1)Temporal models (2)Probabilistic

reasoning

18) Explain the concept of Bayesian network in representing

knowledge in an uncertain domain

UNIT V Short & Big Questions

1) What is machine learning?

A computer program is said to learn from experience E with respect to

some class of tasks T and performance measure P, if its performance at

tasks in T, as measured by P, improves with experience E. Where E is

the training set. T is the task it is made for. P is the performance

measure. 2) What are the components of the learning agent?

(i) Learning element (ii) Performance element (iii) Problem generator 3) Distinguish between supervised and unsupervised learning?

Supervised: Teacher provides training examples & solutions

E.g.

Classification

Unsupervised: No assistance from teacher

E.g. Clustering; Inducing hidden variables

4) What is inductive learning?

If you do have background knowledge, then a question is whether the

learned knowledge is “entailed” by the background knowledge

or not(Entailment can be logical or probabilistic)

If it is entailed, then it is called “deductive” learning

Page 26: question bank-AI-12-13-10144CS601

10144CS601 ARTIFICIAL INTELLIGENCE

QUESTION BANK

26

If it is not entailed, then it is called inductive learning

5) What is clustering in learning?

Given a set of examples, but no labelling of them, group the examples

into “natural” clusters 6) What is the principle of ockham’s razor?

It says that, “entities are not to be multiplied beyond necessity”. 7) Distinguish between classification problem and regression problem in

supervised learning.

In both learning techniques, two quantities are involved. 1. Feature

vector space and the other is the output signal. This can be illustrated

as below:

X1, X2, … Xn Y

If a two dimensional feature vector space is used then it will

correspond to an output quantity. The input may be a pixel value or an

integer representing an employee details in an organisation which

represents an unique value.

If yi is a Boolean, or a member of a discrete set, we will call the problem as a classification problem. When the yi is real valued, the problem is real- valued, we call this a regression problem.

8) What is a decision tree?

A decision tree is a representation in which,

a) Each node is connected to a set of possible answers.

b) Each non leaf node is connected to a test that splits its set of possible

answers into subsets corresponding to different test results. c) Each branch carries a particular test results subset to another node.

9) What is meant by decision network?

For example, the Bayesian network is a directed graph in which each

node constitutes a probability information. Or more elaborately:

(i) A set of random variables makes up the nodes of the network.

Variables may be discrete or continuous.

(ii) A set of directed links or arrows connects pairs of nodes. If there’s

an arrow from node X to Y, X is said to be the parent of Y.

(iii) Each node has a conditional probability distribution

P(Xi| Parents(Xi)) that quantifies the effect of the parents on

the node.

(iv) The graph has no directed cycles(and hence is a directed

acyclic graph, or DAG).

Page 27: question bank-AI-12-13-10144CS601

10144CS601 ARTIFICIAL INTELLIGENCE

QUESTION BANK

27

10) What is a training set? Test set?

While assessing the performance of a learning algorithm, we divide the

feature vectors into test set and training set which are disjoint. We first

apply the learning algorithm on the training set and get a hypothesis.

Now the hypothesis is verified with the test set whether the learner

produces the expected performance. 11) What is noise and overfitting?

In the learning process, if the training set consists of some examples

which will be too hard to establish a crisp hypothesis, the learner rejects

some of them and they are known as noise. Similarly if the feature

vectors are too small to learn a hypothesis, some overfitting will occur.

12) What is cross validation?

It is a technique that is used reduce overfitting. It can be applied to any

learning algorithm not just decision trees. The basic idea is to estimate

how well each hypothesis will predict unseen data. This is done by

setting aside some fraction of data and using it to to test the prediction

performance of a hypothesis induced from the remaining data. K-Fold

cross-validation means k experiments are done. 13) What is mean by split point?

Among the training set data the split-point is the feature vector that

gives the maximum information gain or entropy.

14) What is ensemble learning?

It is to use a collection of hypotheses from the hypothesis space to

make predictions.

15) What is PAC learning?

In computational learning theory, the goal of a learner is to produce a

probably approximately correct (PAC) hypothesis, for a given

approximation (error rate) ϵ and probability . That is, any hypothesis

that is seriously wrong can be found in due course with fewer examples

with a high probability because it will make incorrect predictions. Thus

any hypothesis that is consistent with large set of training sets is

unlikely to be seriously wrong. This is called PAC learning.

16) How to learn decision lists?

Here we are learning a set of logical expressions rather than data. The

first task is to decide whether the input data is learnable. Ultimately a

consistent decision list is returned by an efficient learner. 17) What is the idea behind the current-best-hypothecs search?

Page 28: question bank-AI-12-13-10144CS601

10144CS601 ARTIFICIAL INTELLIGENCE

QUESTION BANK

28

It searches for a consistent hypothesis by suitably applying

specializations /generalizations on the hypothesis. 18) What are version spaces?

The following ideas are used:

Hypothesis are represented by a set of logical sentences.

Incremental construction of hypothesis.

Prior “domain” knowledge can be included/used.

Enables using the full power of logical inference.

19) What do you mean by explanation based learning?

Hypothesis Descriptions ╞═ Classifications

Background ╞═ Hypothesis

General rule follows logically from the background knowledge.

Converting first principle special theories into special purpose knowledge.

In other words, it is a method for extracting general rules from individual

observations. The basic idea behind EBL is first to construct an

explanation of the observation using prior knowledge, and then to

establish a definition of the class of cases for which the same explanation

structure can be used. 20) What is meant by relevance based learning?

In EBL, Background ╞═ Hypothesis. Instead, in RBL, the background

did not completely produce the hypothesis but in addition to

background knowledge, the descriptions and old hypothesis are added

which became deductive in nature. Thus as a result it becomes

Hypothesis Descriptions Background ╞═ Classifications 21) Define Inductive Logic Programming(ILP)

Given an encoding of the known background knowledge and a set of

examples represented as a logical database of facts, an ILP system will

derive a hypothesised logic program which entails all the positive and

none of the negative examples.

Schema:

positive examples + background knowledge => hypothesis.

Inductive logic programming is particularly useful in bioinformatics

and natural language processing. 22) List some of the statistical learning methods available.

Bayesian networks, neural networks and kernel machines 23) What is meant by a neural network?

Page 29: question bank-AI-12-13-10144CS601

10144CS601 ARTIFICIAL INTELLIGENCE

QUESTION BANK

29

It is a mathematical model of neurons(it is a cell in the brain) – devised by

McCulloch and Pitts(1943).

24) Define activation function.

It is used by each node to combine the separate influences received on

its input links into an overall influence. 25) Define threshold function.

It is through this function one simple activation function simply passes

the sum of the input values to determine the node‟s output. 26) Name the different parts of a neural network 27) What is a MAP hypothesis?

In science, the predictions are based on a single most probable hypothesis

– that is an hi that maximises the P(hi| d). This is called maximum a

posteriori or MAP hypothesis.

28) What are the drawbacks of EM algorithm?

It uses much less parameters for learning. Yet learning is harder

because, values of hidden variables are unknown. Also it is an

unsupervised clustering method which is more complicated to perform. 29) True/False

a. Neural networks can be set up to output only 0/1 values.

b. Decision trees with k internal nodes can express any Boolean function

of k Boolean attributes.

c. The human brain can be described, to a first approximation, as a

very large multi-layer feed-forward neural network. d. Any decision tree with Boolean attributes can be converted into an

equivalent feed-forward neural network.

Answer: a)T b) F c) F a feed forward has no internal state and hence no

memory. d) True; a neural net with enough hidden nodes can represent

any Boolean function.

Big Questions 1) Explain the concept of learning using decision trees. 16 2) List few applications of neural networks. 6 3) What is meant by knowledge in learning? 6

4) Explain the role of knowledge in learning with suitable

examples.

16

5) How to represent experience using learning techniques? 6 6) Describe the steps involved in EBL – Explanation Based

Learning method.

8

Page 30: question bank-AI-12-13-10144CS601

10144CS601 ARTIFICIAL INTELLIGENCE

QUESTION BANK

30

7) Explain Inductive Logic Programming using the FOIL

algorithm for the kinship domain.

12

8) Describe inductive learning with inverse deduction. 6 9) Explain the concept of learning using relevance information 16

10) Explain the perceptron – gradient decent learning algorithm. 8 11) Give the advantages of Back-propagation algorithm for learning

in multilayer neural network.

10

12) Write the uses of kernel machines. 4 13) Explain the statistical learning methods 16 14) Explain Q-Learning in detail 15) Describe both passive and active reinforcement learning in

detail

16