M.C.A. (Online Mode) Semester I Examination

Paper Code: OCA-510 MJ
Title: Artificial Intelligence
Time: 3 Hours
Maximum Marks: 70

Q1) Multiple Choice Questions with reasoning — answer any 25 of 30 [25 × 2 = 50]

For every attempted question, select one option and give a brief reason.

No.	Question	A	B	C	D
1	A rational agent selects an action expected to maximise its:	Sensor count	Performance measure	Program length	Knowledge-base size
2	PEAS stands for:	Performance measure, Environment, Actuators, Sensors	Program, Events, Actions, State	Performance, Evaluation, Agents, Search	Percepts, Environment, Actions, State
3	A fully observable environment provides:	Complete relevant state information	Deterministic outcomes	Only one agent	No state changes
4	A stochastic environment has:	Uncertain action outcomes	Complete observability	No dynamics	One possible action
5	A model-based reflex agent maintains:	Internal state	A complete plan only	No memory	Only neural weights
6	In PEAS, actuators allow the agent to:	Sense	Act on the environment	Define utility	Store beliefs
7	BFS expands the shallowest node using a:	LIFO stack	FIFO queue	Heuristic-only queue	Random set
8	UCS expands the node with smallest:	Depth	Path cost g(n)	Heuristic h(n)	Branching factor
9	A* evaluates nodes using:	g(n)+h(n)	g(n)−h(n)	h(n) only	Depth only
10	An admissible heuristic never:	Underestimates	Overestimates true remaining cost	Equals zero	Uses knowledge
11	Alpha–beta pruning changes:	Minimax's final value	Nodes evaluated, not the result	Player order	Utility values
12	A CSP consists of variables, domains, and:	Constraints	Rewards	Search depths	Probabilities
13	Propositional logic represents:	Truth-valued atomic statements	Quantified objects	Probability tables	Neural weights
14	First-order logic adds:	Only truth tables	Predicates and quantifiers	Probabilities	Search costs
15	Modus ponens derives Q from P and:	Q→P	P→Q	¬P	P∨Q
16	Propositional resolution operates on:	Clauses	Frames only	Probabilities	Weights
17	A semantic network mainly uses:	Nodes and labelled links	A numeric matrix	A queue	Source code
18	Backward chaining begins with:	All facts	A goal to prove	Random actions	A probability prior
19	Bayes' theorem computes a posterior from:	Likelihood and prior	Variance only	Search cost	Entropy only
20	Independent A and B satisfy:	P(A∩B)=P(A)P(B)	P(A∩B)=P(A)+P(B)	P(A|B)=0	P(A)=P(B)
21	A Bayesian network is a:	Directed acyclic graph with conditional tables	Flat truth table	Search tree	Rule list only
22	Naive Bayes assumes features are:	Conditionally independent given the class	Always equally likely	Mutually exclusive	Unobserved
23	An HMM contains:	Hidden states and observable emissions	Only visible states	No transition probabilities	A search heuristic
24	Under uncertainty, a rational choice maximises:	Expected utility	Best possible outcome alone	Immediate reward only	Number of outcomes
25	A perceptron computes:	A weighted sum followed by activation	A database join	A shortest path	A truth table only
26	ReLU is commonly defined as:	max(0,x)	1/(1+e^-x)	x²	sign(x)
27	A single linear perceptron cannot represent:	AND	OR	XOR	NOT
28	Backpropagation uses the chain rule to compute:	Gradients	Search paths	Priors	Logical clauses
29	Gradient descent updates a weight generally in the direction:	Opposite the loss gradient	Along the loss gradient	Of maximum activation	Of random increase
30	Softmax converts output scores into:	A class-probability distribution	Binary inputs	Hidden states	Search costs

Q2) Answer any five in approximately two sentences [5 × 2 = 10]

1. Distinguish rationality from omniscience in an intelligent agent.
2. Define a search problem using its main components.
3. Distinguish greedy best-first search from A*.
4. What do MRV and forward checking do in a CSP?
5. Distinguish a semantic network from a frame.
6. Define conditional independence in a Bayesian network.
7. State the perceptron learning rule.

Q3) Answer any two in approximately ten lines [2 × 5 = 10]

1. Compare BFS, DFS, UCS, greedy best-first search, and A* by selection rule, completeness, and optimality.
2. Explain propositional resolution by converting P→Q, P, and ¬Q to clauses and deriving a contradiction.
3. A disease has prior probability 0.01. A test has sensitivity 0.90 and false-positive rate 0.05. Find P(Disease | Positive).
4. For input x=(1,0), target t=1, initial weights (0.2,−0.1), bias −0.3, learning rate 0.1, and step output 1 when net ≥0 otherwise 0, perform one perceptron update.