UGC NET Computer Science Artificial Intelligence Previous Year Questions (PYQs)

All 133 Artificial Intelligence PYQs

Ordered newest exam first. Each question links back to the full paper it came from.

1. Question 1

   UGC NET June 2026

   Which field deals with interaction between computers and humans using natural language?

   1. APROLOG
   2. BELIZA
   3. CNLP
   4. DLISP

   Answer: (C) NLP

   Explanation

   Natural Language Processing (NLP) is the AI field concerned with computer processing and generation of human language. PROLOG and LISP are programming languages, while ELIZA is an early NLP-style program rather than the name of the field.

   Sign in to read the full explanation

   Login / Signup

2. Question 2

   UGC NET June 2026

   Classifying email as spam, labeling web pages by context, and voice recognition are examples of which approach?

   1. AReinforcement learning
   2. BShortest-path algorithm
   3. CSupervised learning
   4. DUnsupervised learning

   Answer: (C) Supervised learning

   Explanation

   These tasks learn from examples with known labels, such as spam/not spam or a known spoken word. That is supervised learning; unsupervised learning has no target labels, while reinforcement learning optimises rewards through interaction.

   Sign in to read the full explanation

   Login / Signup

3. Question 3

   UGC NET June 2026

   For the shown neural network, what are the net input yᵢₙ and final threshold output y?

   

   Neural-network inputs, weights, bias, and threshold function.

   1. Ayᵢₙ = 2.03, y = 1
   2. Byᵢₙ = 1.53, y = 1
   3. Cyᵢₙ = 1.07, y = 1
   4. Dyᵢₙ = 0.57, y = 0

   Answer: (C) yᵢₙ = 1.07, y = 1

   Explanation

   Using the diagram values, yᵢₙ = (0.5×0.5) + (0.6×−0.8) + (0.8×1) + 0.5 = 1.07. Since the threshold function returns 1 for an input at least 1, y = 1; the alternatives use an incorrect sign, omit a term, or put the total below the threshold.

   Sign in to read the full explanation

   Login / Signup

4. Question 4

   UGC NET June 2026

   A doctor knows P(s | m) = 0.70, P(m) = 1/50,000, and P(s) = 0.01, where s is a stiff neck and m is meningitis. Compute P(m | s).

   1. A0.014
   2. B0.0014
   3. C0.14
   4. D0.00014

   Answer: (B) 0.0014

   Explanation

   Bayes’ rule gives P(m | s) = P(s | m)P(m)/P(s) = 0.70 × (1/50,000) / 0.01 = 0.0014. The disease remains unlikely even after observing the symptom because its prior probability is very small.

   Sign in to read the full explanation

   Login / Signup

5. Question 5

   UGC NET January 2026

   In a multi-agent system (MAS), what are the key characteristics of individual agents?

   A. They must be identical in their capabilities and goals.
   B. They are autonomous and act independently.
   C. They interact with each other and their environment.
   D. They are controlled by a single, centralized authority at all times.

   1. AA, B & C Only
   2. BB, C & D Only
   3. CB & C Only
   4. DA & D Only

   Answer: (C) B & C Only

   Explanation

   A multi-agent system is defined by agents that operate on their own and communicate, not by uniformity or central control.

   • A. False. Agents in a MAS are typically heterogeneous, with different capabilities and goals suited to their roles.
   • B. True. Autonomy, the ability to act without direct external control, is a defining property of an agent.
   • C. True. Agents interact with each other and with their environment to coordinate or compete toward their goals.
   • D. False. MAS are generally decentralized, which is precisely what distinguishes them from a single centrally controlled system.

   Only B and C hold, so B & C Only is correct.

   Sign in to read the full explanation

   Login / Signup

6. Question 6

   UGC NET January 2026

   Consider the following statements about reinforcement learning.

   A. The adaptive dynamic programming agent leaves the transition model between states utilizes to solve the corresponding Markov decision process using dynamic programming.
   B. Temporal difference needs a transition model to perform its updates.
   C. The prioritized sweeping heuristic focuses on adjusting states with successors that have undergone significant changes in utility estimates.
   D. The approach of modified policy iteration involves adopting a simplified value utility estimates following each change to the learned model.

   1. AA, B & C Only
   2. BA, B & D Only
   3. CA, C & D Only
   4. DB, C & D Only

   Answer: (C) A, C & D Only

   Explanation

   Adaptive dynamic programming (ADP), temporal difference (TD) learning, prioritized sweeping and modified policy iteration are distinguished mainly by whether they use a transition model and how they update value estimates.

   • A. True. An ADP agent learns the transition model of the environment and then solves the resulting Markov decision process using standard dynamic programming.
   • B. False. Temporal difference learning is model-free by design, updating value estimates directly from observed transitions and rewards without ever building a transition model.
   • C. True. Prioritized sweeping directs computation toward states whose successors have recently had large changes in their utility estimates, since updating those states first propagates useful information fastest.
   • D. True. Modified policy iteration replaces full, exact policy evaluation with a small number of simplified value-update passes after each change to the learned model, trading some accuracy for speed.

   A, C and D hold while B misdescribes temporal difference learning as needing a model, so A, C & D Only is correct.

   Sign in to read the full explanation

   Login / Signup

7. Question 7

   UGC NET January 2026

   Which logic is designed to address the problem of referential opacity in propositional attitudes?

   1. ARegular Logic
   2. BModal Logic
   3. CFirst-order logic
   4. DPredicate logic

   Answer: (B) Modal Logic

   Explanation

   Referential opacity arises in propositional-attitude contexts such as belief, knowledge, and desire, where substituting equal terms inside the attitude can change the truth value of the whole statement, unlike in standard extensional logic.

   Modal logic extends classical logic with operators for necessity and possibility, and its extensions such as epistemic and doxastic logic are specifically built to represent propositional attitudes like belief and knowledge without collapsing under substitution of co-referring terms. This makes modal logic the framework designed to handle referential opacity.

   First-order logic and predicate logic are extensional systems, so they cannot correctly represent attitude contexts where substitution of equals fails, and there is no established logic called Regular Logic that addresses this problem.

   Sign in to read the full explanation

   Login / Signup

8. Question 8

   UGC NET January 2026

   Arrange the following sequence related to the FASTUS system.
   A. Basic-group handling
   B. Structure merging
   C. Tokenization
   D. Complex phrase handling
   E. Complex word handling

   1. AC, A, D, E, B
   2. BE, A, C, D, B
   3. CC, E, A, D, B
   4. DC, E, D, A, B

   Answer: (C) C, E, A, D, B

   Explanation

   FASTUS is a cascaded finite-state system for information extraction, and each stage builds on the output of the previous one.

   C, E, A, D, B.

   • C. Tokenization — breaks the raw text into individual tokens, the first step.
   • E. Complex word handling — recognizes multi-word expressions, proper names, and other fixed word groups from the tokens.
   • A. Basic-group handling — identifies simple noun and verb groups from the processed words.
   • D. Complex phrase handling — combines the basic groups into larger noun and verb phrases.
   • B. Structure merging — the final stage, merging the recognized phrases into complete domain-level event structures.

   Sign in to read the full explanation

   Login / Signup

9. Question 9

   UGC NET January 2026

   Match List-I (AI subfields) with List-II.

   LIST-I	LIST-II
   A. Expert systems	I. Medical Diagnosis
   B. NLP	II. Text Summarization
   C. Computer Vision	III. Image Classification
   D. ML	IV. Predictive Analysis

   1. AA-I, B-II, C-III, D-IV
   2. BA-II, B-I, C-III, D-IV
   3. CA-II, B-III, C-I, D-IV
   4. DA-III, B-II, C-I, D-IV

   Answer: (A) A-I, B-II, C-III, D-IV

   Explanation

   Each pairing follows from the primary application area of each AI subfield.

   A-I, B-II, C-III, D-IV.

   • A. Expert systems — I. Medical Diagnosis. Expert systems apply encoded domain rules to specific problems, and medical diagnosis (as in systems like MYCIN) is one of the earliest and most famous applications.
   • B. NLP — II. Text Summarization. Natural language processing works directly with text, making summarization a natural fit.
   • C. Computer Vision — III. Image Classification. Computer vision analyzes and interprets images, and image classification is one of its core tasks.
   • D. Machine Learning — IV. Predictive Analysis. Machine learning builds models from historical data to forecast outcomes, which is the basis of predictive analysis.

   Sign in to read the full explanation

   Login / Signup

10. Question 10

    UGC NET January 2026

    Assertion (A): Unsupervised learning algorithms are used for tasks like data summarization and exploratory data analysis.
    Reason (R): Unsupervised learning requires a labelled dataset to find relationships and patterns in data.

    1. ABoth A and R are correct and R is the correct explanation of A
    2. BBoth A and R are correct but R is NOT the correct explanation of A
    3. CA is correct but R is not correct
    4. DA is not correct but R is correct

    Answer: (C) A is correct but R is not correct

    Explanation

    Unsupervised learning algorithms, such as clustering and dimensionality reduction, are commonly applied to summarize data and explore its structure without needing predefined output labels, so Assertion A is correct.

    Reason R claims the opposite of how unsupervised learning is defined. Unsupervised learning specifically works on unlabelled data to discover hidden patterns and structure, while it is supervised learning that requires a labelled dataset. So R is incorrect.

    A is correct but R is not correct.

    Sign in to read the full explanation

    Login / Signup

11. Question 11

    UGC NET January 2026

    Which planning approach is used for an environment with no observations?

    1. AHierarchical planning
    2. BReactive planning
    3. CConformant planning
    4. DContingency Planning

    Answer: (C) Conformant planning

    Explanation

    Conformant planning is designed for environments where the agent has no sensing capability at all, so it cannot observe the current state during execution.

    Because no observations are available, a conformant plan must be built in advance as a fixed sequence of actions that is guaranteed to reach the goal regardless of the actual initial state or outcome, with no branching based on feedback.

    • A. Hierarchical planning — decomposes a task into abstract subtasks, unrelated to observability.
    • B. Reactive planning — selects actions directly from current percepts without deliberative lookahead, which still needs some observation.
    • C. Conformant planning — correct. It works with zero observability by committing to a fixed action sequence in advance.
    • D. Contingency planning — assumes partial observability and builds a plan with branches selected using sensed information.

    Sign in to read the full explanation

    Login / Signup

12. Question 12

    UGC NET January 2026

    Arrange the following steps of the Natural Language Processing pipeline in the correct order.

    A. Syntax Analysis
    B. Morphological Analysis
    C. Pragmatics
    D. Semantic Analysis

    1. AA, B, C, D
    2. BB, A, D, C
    3. CB, D, A, C
    4. DA, C, B, D

    Answer: (B) B, A, D, C

    Explanation

    A standard NLP pipeline processes text in stages that build on each other, starting from the smallest linguistic unit and moving toward context.

    Morphological analysis (B) comes first and breaks words into roots and affixes.
    Syntax analysis (A) follows and checks how words combine grammatically into sentence structure.
    Semantic analysis (D) comes next and interprets the meaning of the syntactically valid structure.
    Pragmatics (C) is the final stage and interprets meaning in context, including intent and real-world knowledge beyond the literal sentence.

    This gives the order B, A, D, C.

    Sign in to read the full explanation

    Login / Signup

13. Question 13

    UGC NET January 2026

    In artificial intelligence, the term "Turing Test" is primarily associated with:

    1. ATesting neural network convergence
    2. BMeasuring computational efficiency of an algorithm
    3. CEvaluating a machine's ability to exhibit human-like behavior
    4. DVerifying correctness of an AI search algorithm

    Answer: (C) Evaluating a machine's ability to exhibit human-like behavior

    Explanation

    The Turing Test, proposed by Alan Turing, asks whether a human evaluator can distinguish a machine's text responses from a human's responses in conversation.

    It does not measure raw computational speed, algorithmic correctness, or neural network training behavior. It specifically evaluates whether a machine can exhibit behavior indistinguishable from human intelligence.

    So the Turing Test is about evaluating a machine's ability to exhibit human-like behavior.

    Sign in to read the full explanation

    Login / Signup

14. Question 14

    UGC NET January 2026

    Assertion (A): Neural networks are capable of approximating any continuous function to an arbitrary degree of accuracy.
    Reason (R): This is explained by the universal approximation theorem.

    1. ABoth A and R are correct and R is the correct explanation of A
    2. BBoth A and R are correct but R is NOT the correct explanation of A
    3. CA is correct but R is not correct
    4. DA is not correct but R is correct

    Answer: (A) Both A and R are correct and R is the correct explanation of A

    Explanation

    The universal approximation theorem states that a feedforward neural network with at least one hidden layer of sufficient width and a suitable non-linear activation function can approximate any continuous function on a compact domain to any desired degree of accuracy.

    Assertion A is a direct statement of this capability, and Reason R correctly names the theorem that explains it.

    Since R is precisely the theoretical justification for A, R is the correct explanation of A.

    Sign in to read the full explanation

    Login / Signup

15. Question 15

    UGC NET January 2026

    Match List-I (Horn-clause terms) with List-II.

    List-I	List-II
    A. Definite clause	I. Forward chaining
    B. Horn clause	II. At most one is positive literal
    C. Goal clause	III. No positive literals
    D. Inference with Horn clause	IV. Exactly one is positive literal

    1. AA-II, B-IV, C-III, D-I
    2. BA-IV, B-III, C-I, D-III
    3. CA-II, B-III, C-I, D-IV
    4. DA-IV, B-II, C-III, D-I

    Answer: (D) A-IV, B-II, C-III, D-I

    Explanation

    Horn-clause logic classifies clauses by how many positive literals they contain.

    A definite clause has exactly one positive literal, so A pairs with IV.
    A Horn clause has at most one positive literal, a broader category that includes definite clauses, so B pairs with II.
    A goal clause is a Horn clause with no positive literals at all, so C pairs with III.
    The standard inference procedure used with Horn clauses is forward chaining, so D pairs with I.

    This gives A-IV, B-II, C-III, D-I.

    Sign in to read the full explanation

    Login / Signup

16. Question 16

    UGC NET January 2026

    Assertion (A): In supervised learning, the model is trained using labelled data.
    Reason (R): Supervised learning algorithms find patterns in data to predict outcomes without any prior knowledge of the correct output.

    1. ABoth A and R are correct and R is the correct explanation of A
    2. BBoth A and R are correct but R is NOT the correct explanation of A
    3. CA is correct but R is not correct
    4. DA is not correct but R is correct

    Answer: (C) A is correct but R is not correct

    Explanation

    Supervised learning trains a model on a dataset where every input is paired with a known, correct output label, so Assertion A is correct.

    Reason R is not correct. It describes finding patterns without any prior knowledge of the correct output, which is the definition of unsupervised learning, not supervised learning. Supervised learning is defined by the opposite property, since it explicitly uses the correct output labels during training to guide the model toward accurate predictions.

    Since A is true and R actually describes a different learning paradigm, the correct choice is that A is correct but R is not correct.

    Sign in to read the full explanation

    Login / Signup

17. Question 17

    UGC NET June 2025

    Which is not a valid property of λ-cuts of fuzzy relations R̃ and S̃?

    1. A(R̃ ∪ S̃)λ = Rλ ∪ Sλ
    2. B(R̃ ∩ S̃)λ = Rλ ∩ Sλ
    3. CThe λ-cut of the complement of R̃ differs from the complement of Rλ except when λ = 1.
    4. DFor λ ≤ β with 0 ≤ β ≤ 1, Rβ ⊆ Rλ.

    Answer: (C) The λ-cut of the complement of R̃ differs from the complement of Rλ except when λ = 1.

    Explanation

    Union and intersection commute with λ-cuts, and increasing the threshold can only shrink a cut, so Rβ ⊆ Rλ for β ≥ λ. Complement does not have the stated exceptional equality at λ = 1: its cut is governed by the complementary membership threshold, not simply by taking the ordinary complement of Rλ.

    Sign in to read the full explanation

    Login / Signup

18. Question 18

    UGC NET June 2025

    Which is not a parent-selection technique in genetic-algorithm implementations?

    1. ARadial
    2. BTournament
    3. CBoltzmann
    4. DRank

    Answer: (A) Radial

    Explanation

    Tournament, rank-based, and Boltzmann selection are ways to choose parents according to fitness. “Radial” is not a standard parent-selection method; it is more commonly associated with radial-basis functions.

    Sign in to read the full explanation

    Login / Signup

19. Question 19

    UGC NET June 2025

    With η as learning rate and xi as the ith input, which update rule best describes a Kohonen self-organizing map weight wij?

    1. Awij(new) = (1 − η)wij(old) + ηxi
    2. Bwij(new) = ηwij(old) + (1 − η)xi
    3. Cwij(new) = (1 − η)wij(old) + xi
    4. Dwij(new) = wij(old) + ηxi

    Answer: (A) wij(new) = (1 − η)wij(old) + ηxi

    Explanation

    A SOM moves a winning neuron's weight vector a fraction η toward the input: w(new) = w(old) + η(x − w(old)). Rearranging gives (1 − η)w(old) + ηx, so the new weight is a convex combination of the old weight and input.

    Sign in to read the full explanation

    Login / Signup

20. Question 20

    UGC NET June 2025

    After applying samples S1 through S4 in order to a perceptron with initial weights w1 = w2 = 0, learning rate 0.1, no bias, and output 0 when yin = 0, what are the updated weights?

    Sample	x1	x2	Target
    S1	0	0	0
    S2	0	1	1
    S3	1	0	1
    S4	1	1	1

    1. Aw1 = 0.1, w2 = 0.1
    2. Bw1 = 0.0, w2 = 0.2
    3. Cw1 = 0.0, w2 = 0.1
    4. Dw1 = 0.2, w2 = 0.2

    Answer: (A) w1 = 0.1, w2 = 0.1

    Explanation

    S1 has zero inputs and target 0, so it changes nothing. S2 has input (0,1), target 1, and adds 0.1 to w2; S3 similarly adds 0.1 to w1. For S4, yin = 0.2 and the observed output is already 1, matching its target, so no further update occurs.

    Sign in to read the full explanation

    Login / Signup

21. Question 21

    UGC NET June 2025

    For knowledge-base design, arrange choosing a task domain, selecting atoms for propositions, axiomatizing the domain, and asking questions about the intended interpretation.

    1. AC → A → D → B
    2. BC → A → B → D
    3. CB → C → A → D
    4. DA → C → B → D

    Answer: (A) C → A → D → B

    Explanation

    The task domain sets the scope. The designer then chooses the propositional vocabulary, writes axioms in that vocabulary, and can ask whether the resulting representation has the intended interpretation. This produces C, A, D, B.

    Sign in to read the full explanation

    Login / Signup

22. Question 22

    UGC NET June 2025

    Arrange these genetic-algorithm actions: calculate population fitness, select parents, apply crossover, and mutate each locus.

    1. AA → C → B → D
    2. BC → A → D → B
    3. CC → A → B → D
    4. DA → D → B → C

    Answer: (B) C → A → D → B

    Explanation

    Fitness must be computed before fitness-informed parent selection. Selected parents undergo crossover to create offspring, and mutation is then applied probabilistically at loci. The standard order is C, A, D, B.

    Sign in to read the full explanation

    Login / Signup

23. Question 23

    UGC NET June 2025

    Which statements about agent systems are correct: an agent system comprises an agent and its environment; an agent controller receives percepts and sends commands; actuators are always reliable; and actuators convert stimuli into percepts?

    1. AA and B only
    2. BB and D only
    3. CC and D only
    4. DB, C, and D only

    Answer: (A) A and B only

    Explanation

    An agent acts in an environment and its controller maps percepts to actions. Sensors, not actuators, convert environmental stimuli to percepts; actuators execute actions and can be noisy or unreliable. Thus only A and B are correct.

    Sign in to read the full explanation

    Login / Signup

24. Question 24

    UGC NET June 2025

    Which STRIPS statements are correct: it is feature-centric; world-state features are primitive and derived; an action has preconditions and effects; and it directly defines conditional effects?

    1. AA and C only
    2. BB and C only
    3. CA and D only
    4. DB, C, and D only

    Answer: (B) B and C only

    Explanation

    STRIPS represents a state using features or propositions, commonly distinguished as primitive and derived, and describes actions through preconditions and effects. Basic STRIPS effects are unconditional, so it does not directly express conditional effects. Therefore B and C are the correct pair.

    Sign in to read the full explanation

    Login / Signup

25. Question 25

    UGC NET June 2025

    Match fuzzy-set intersection, bounded sum, bounded difference, and algebraic sum with their membership functions.

    Operation	Membership function
    A. Intersection	I. μA(x) + μB(x) − μA(x)μB(x)
    B. Bounded Sum	II. max(0, μA(x) − μB(x))
    C. Bounded Difference	III. min(1, μA(x) + μB(x))
    D. Algebraic Sum	IV. min(μA(x), μB(x))

    1. AA-III, B-I, C-II, D-IV
    2. BA-III, B-II, C-IV, D-I
    3. CA-IV, B-III, C-II, D-I
    4. DA-II, B-I, C-III, D-IV

    Answer: (C) A-IV, B-III, C-II, D-I

    Explanation

    Intersection takes the smaller membership, min(μA, μB). Bounded sum caps ordinary addition at 1, while bounded difference floors subtraction at 0. Algebraic sum is the probabilistic-style formula μA + μB − μAμB, not simple capped addition. Therefore A-IV, B-III, C-II, D-I.

    Sign in to read the full explanation

    Login / Signup
26. Practise Artificial Intelligence in a timed set

27. Question 26

    UGC NET June 2025

    Match worst individual, best individual, sum of fitness, and median fitness termination criteria with their convergence descriptions.

    Criterion	Description
    A. Worst Individual	I. At least half the individuals are at or above convergence value
    B. Best Individual	II. Virtually all individuals are within a fitness range
    C. Sum of Fitness	III. Entire population meets a minimum standard
    D. Median Fitness	IV. At least one best solution / faster convergence

    1. AA-III, B-IV, C-II, D-I
    2. BA-IV, B-III, C-I, D-II
    3. CA-II, B-I, C-III, D-IV
    4. DA-II, B-III, C-IV, D-I

    Answer: (A) A-III, B-IV, C-II, D-I

    Explanation

    A worst-individual threshold ensures the entire population meets a minimum standard. A best-individual criterion guarantees at least one best solution, sum of fitness constrains the population to a fitness range, and median fitness means at least half the population is at or above the convergence value. Thus A-III, B-IV, C-II, D-I.

    Sign in to read the full explanation

    Login / Signup

28. Question 27

    UGC NET January 2025

    What is the correct sequence of steps used in knowledge-base design?

    A. Ask questions about the intended interpretation.
    B. Choose the task domain or world to represent.
    C. Select atoms to represent propositions of interest.
    D. Tell the system propositions that are true in the intended interpretation.

    1. AA, B, C, D
    2. BB, C, D, A
    3. CC, D, A, B
    4. DD, A, B, C

    Answer: (B) B, C, D, A

    Explanation

    Knowledge engineering starts by identifying the task and world, then selecting vocabulary, axiomatizing known facts, and finally asking questions to test the intended interpretation. The correct option follows from the stated definition or calculation; the other options omit a necessary condition or describe a different concept.

    Sign in to read the full explanation

    Login / Signup

29. Question 28

    UGC NET January 2025

    Which are correct for a neural network?

    A. Training time depends on network size.
    B. A neural network can be simulated on a conventional computer.
    C. A neural network mimics the way the human brain works.
    D. A neural network includes feedback.

    1. AA and B only
    2. BA, C and D only
    3. CA, B and C only
    4. DA and C only

    Answer: (B) A, C and D only

    Explanation

    Training cost grows with model size, networks can run on conventional computers, and they are brain-inspired abstractions. Feedback is not present in every neural-network architecture. The correct option follows from the stated definition or calculation; the other options omit a necessary condition or describe a different concept.

    Sign in to read the full explanation

    Login / Signup

30. Question 29

    UGC NET January 2025

    Which is not a component of the classic planning definition?

    1. AInit
    2. BDomain
    3. CAction
    4. DGoal

    Answer: (B) Domain

    Explanation

    A classical planning problem specifies an initial state, actions and a goal. The domain supplies the model but is not one of the basic components named in this formulation. The correct option follows from the stated definition or calculation; the other options omit a necessary condition or describe a different concept.

    Sign in to read the full explanation

    Login / Signup

31. Question 30

    UGC NET January 2025

    Which one of the following is not related to feed-forward networks in the backpropagation algorithm?

    1. ABoolean function
    2. BContinuous function
    3. CArbitrary function
    4. DGreedy function

    Answer: (D) Greedy function

    Explanation

    Feed-forward neural networks are commonly discussed as function approximators for Boolean, continuous and arbitrary functions. “Greedy function” is not such a network-function category. The correct option follows from the stated definition or calculation; the other options omit a necessary condition or describe a different concept.

    Sign in to read the full explanation

    Login / Signup

32. Question 31

    UGC NET January 2025

    Definitions organised into the four categories Thinking Humanly, Thinking Rationally, Acting Humanly and Acting Rationally describe

    1. AMachine Learning
    2. BDeep Learning
    3. CArtificial Intelligence
    4. DNeural Network

    Answer: (C) Artificial Intelligence

    Explanation

    These four viewpoints are the classic framework used to define artificial intelligence: human-like versus rational thinking and acting. The correct option follows from the stated definition or calculation; the other options omit a necessary condition or describe a different concept.

    Sign in to read the full explanation

    Login / Signup

33. Question 32

    UGC NET January 2025

    Match List-I with List-II.

    List-I	List-II
    A. Decision tree B. Supervised learning C. Artificial neural network D. Instance-based learning	I. Delta learning rule II. Self-organizing map III. C4.5 algorithm IV. Nonlinear regression algorithm

    1. AA-I, B-II, C-III, D-IV
    2. BA-II, B-III, C-IV, D-I
    3. CA-III, B-IV, C-I, D-II
    4. DA-IV, B-I, C-II, D-III

    Answer: (C) A-III, B-IV, C-I, D-II

    Explanation

    Decision trees use C4.5, supervised learning includes nonlinear regression, artificial neural networks use the delta learning rule, and instance-based learning includes self-organizing maps in the supplied matching.

    Sign in to read the full explanation

    Login / Signup

34. Question 33

    UGC NET August 2024

    Match List-I with List-II.

    List-I	List-II
    (A) The activation function	(I) is called the delta rule.
    (B) The learning method of perceptron	(II) is one of the key components of the perceptron, as in the most common neural network architecture.
    (C) Areas of application of artificial neural networks	(III) is always boolean, like a switch.
    (D) The output of the perceptron	(IV) system identification and control.

    1. A(A)-(II), (B)-(IV), (C)-(III), (D)-(I)
    2. B(A)-(IV), (B)-(III), (C)-(II), (D)-(I)
    3. C(A)-(II), (B)-(I), (C)-(IV), (D)-(III)
    4. D(A)-(III), (B)-(IV), (C)-(II), (D)-(I)

    Answer: (C) (A)-(II), (B)-(I), (C)-(IV), (D)-(III)

    Explanation

    The activation function is one of the key components of a perceptron, so (A)-(II). A perceptron learns by the delta rule, giving (B)-(I). Artificial neural networks are applied to tasks such as system identification and control, so (C)-(IV). A perceptron emits a boolean, switch-like output, so (D)-(III). Only option C carries all four pairings.

    Sign in to read the full explanation

    Login / Signup

35. Question 34

    UGC NET August 2024

    Arrange the following steps in proper sequence
    involved in a Genetic Algorithm :
    A. Selection
    B. Initialization
    C. Crossover
    D. Mutation
    E. Evaluation

    1. A(A), (B), (C), (D), (E)
    2. B(E), (A), (B), (D), (C)
    3. C(B), (E), (A), (C), (D)
    4. D(A), (C), (B), (D), (E)

    Answer: (C) (B), (E), (A), (C), (D)

    Explanation

    A genetic algorithm initialises a population (B), evaluates each individual’s fitness (E), selects parents according to that fitness (A), recombines them by crossover (C), and finally perturbs the offspring by mutation (D). Selection needs fitness scores, so evaluation must come before it. That order is option C.

    Sign in to read the full explanation

    Login / Signup

36. Question 35

    UGC NET August 2024

    Arrange the following steps in a proper sequence
    for the process of training a neural network.
    A. Weight initialization
    B. Feed forward
    C. Back Propagation
    D. Loss Calculation
    E. Weight Update

    1. A(A), (B), (D), (C), (E)
    2. B(D), (B), (A), (C), (E)
    3. C(A), (C), (D), (B), (E)
    4. D(E), (C), (B), (D), (A)

    Answer: (A) (A), (B), (D), (C), (E)

    Explanation

    Training runs: initialise the weights (A), feed the input forward to obtain predictions (B), compute the loss against the targets (D), back-propagate the gradients (C), then update the weights (E). The loss must exist before gradients can be taken, so (D) precedes (C). That order is option A.

    Sign in to read the full explanation

    Login / Signup

37. Question 36

    UGC NET August 2024

    Match List-I with List-II.

    List-I	List-II
    (A) Natural language processing	(I) A method of training algorithm by rewarding desired behaviour and/or punishing undesired one.
    (B) Reinforcement learning	(II) System designed to emulate the decision making abilities of a human expert.
    (C) Support vector machine	(III) A branch of AI focused on understanding and generating human language.
    (D) Expert system	(IV) A machine learning technique that finds the hyperplane that best separates different classes in a feature space.

    1. A(A)-(I), (B)-(II), (C)-(IV), (D)-(III)
    2. B(A)-(III), (B)-(II), (C)-(I), (D)-(IV)
    3. C(A)-(III), (B)-(I), (C)-(IV), (D)-(II)
    4. D(A)-(II), (B)-(IV), (C)-(III), (D)-(I)

    Answer: (C) (A)-(III), (B)-(I), (C)-(IV), (D)-(II)

    Explanation

    Natural language processing is the AI branch for understanding and generating human language, so (A)-(III). Reinforcement learning trains by rewarding desired and punishing undesired behaviour, giving (B)-(I). A support vector machine finds the separating hyperplane, so (C)-(IV). An expert system emulates a human expert’s decisions, so (D)-(II). Option C is correct.

    Sign in to read the full explanation

    Login / Signup

38. Question 37

    UGC NET August 2024

    What is the role of Back Propagation
    Algorithm?

    Artificial Neural Networks (ANNs) are computational models inspired by the human brain’s neural networks. They consist of interconnected nodes, or neurons, organized into layers: an input layer, one or more hidden layers and an output layer. Each connection between neurons has a weight that adjusts as learning progresses, allowing the network to adapt and improve its performance. ANNs are particularly effective in recognizing patterns, making them valuable for tasks such as image and speech recognition, natural language processing and predictive analytics. Learning in ANNs typically involves training algorithms like back propagation, which minimize the error by adjusting the weights. As a subset of machine learning, ANNs have revolutionized the field of Artificial Intelligence by providing solutions to complex problems that traditional algorithms struggle with.

    1. ATo reduce error
    2. BTo secure network
    3. CTo control speed of data
    4. DTo add different layers

    Answer: (A) To reduce error

    Explanation

    Back-propagation computes the gradient of the loss with respect to every weight by applying the chain rule backwards through the network, so that the weights can then be adjusted to reduce the error. Its purpose is error reduction, not securing the network, controlling data speed or adding layers. Option A is correct.

    Sign in to read the full explanation

    Login / Signup

39. Question 38

    UGC NET August 2024

    What is the role of weights in an ANN ?

    Artificial Neural Networks (ANNs) are computational models inspired by the human brain’s neural networks. They consist of interconnected nodes, or neurons, organized into layers: an input layer, one or more hidden layers and an output layer. Each connection between neurons has a weight that adjusts as learning progresses, allowing the network to adapt and improve its performance. ANNs are particularly effective in recognizing patterns, making them valuable for tasks such as image and speech recognition, natural language processing and predictive analytics. Learning in ANNs typically involves training algorithms like back propagation, which minimize the error by adjusting the weights. As a subset of machine learning, ANNs have revolutionized the field of Artificial Intelligence by providing solutions to complex problems that traditional algorithms struggle with.

    1. ATo store data
    2. BTo adjust and improve network performance
    3. CTo control the speed
    4. DTo secure the network

    Answer: (B) To adjust and improve network performance

    Explanation

    The weight on each connection scales the signal passing along it, and learning consists precisely of adjusting these weights so the network’s predictions improve. They are trainable parameters, not a data store, a speed control or a security mechanism. Option B is correct.

    Sign in to read the full explanation

    Login / Signup

40. Question 39

    UGC NET August 2024

    Which of the following is/are the
    application area(s) of ANN ?
    A. Natural Language Processing
    B. Image Processing
    C. Pattern Recognition
    D. Speech Recognition
    Choose the correct answer from the
    options given below :

    Artificial Neural Networks (ANNs) are computational models inspired by the human brain’s neural networks. They consist of interconnected nodes, or neurons, organized into layers: an input layer, one or more hidden layers and an output layer. Each connection between neurons has a weight that adjusts as learning progresses, allowing the network to adapt and improve its performance. ANNs are particularly effective in recognizing patterns, making them valuable for tasks such as image and speech recognition, natural language processing and predictive analytics. Learning in ANNs typically involves training algorithms like back propagation, which minimize the error by adjusting the weights. As a subset of machine learning, ANNs have revolutionized the field of Artificial Intelligence by providing solutions to complex problems that traditional algorithms struggle with.

    1. A(A) and (B) Only
    2. B(B) and (C) Only
    3. C(A), (B) and (C) Only
    4. D(A), (B), (C) and (D)

    Answer: (D) (A), (B), (C) and (D)

    Explanation

    The passage lists pattern recognition together with image recognition, speech recognition, natural language processing and predictive analytics among the tasks artificial neural networks handle well. All four listed areas therefore apply, so option D is correct.

    Sign in to read the full explanation

    Login / Signup

41. Question 40

    UGC NET August 2024

    Artificial Neural Networks (ANNs) are
    inspired by :

    Artificial Neural Networks (ANNs) are computational models inspired by the human brain’s neural networks. They consist of interconnected nodes, or neurons, organized into layers: an input layer, one or more hidden layers and an output layer. Each connection between neurons has a weight that adjusts as learning progresses, allowing the network to adapt and improve its performance. ANNs are particularly effective in recognizing patterns, making them valuable for tasks such as image and speech recognition, natural language processing and predictive analytics. Learning in ANNs typically involves training algorithms like back propagation, which minimize the error by adjusting the weights. As a subset of machine learning, ANNs have revolutionized the field of Artificial Intelligence by providing solutions to complex problems that traditional algorithms struggle with.

    1. AQuantum mechanics
    2. BHuman brain's neural network
    3. CComputer Hardware architecture
    4. DGenetic algorithm

    Answer: (B) Human brain's neural network

    Explanation

    Artificial neural networks are computational models inspired by the human brain’s neural networks: interconnected nodes, or neurons, arranged in layers with weighted connections that adjust as learning progresses. Option B is correct.

    Sign in to read the full explanation

    Login / Signup

42. Question 41

    UGC NET August 2024

    Which of the following layers may be more
    than one in number ?

    Artificial Neural Networks (ANNs) are computational models inspired by the human brain’s neural networks. They consist of interconnected nodes, or neurons, organized into layers: an input layer, one or more hidden layers and an output layer. Each connection between neurons has a weight that adjusts as learning progresses, allowing the network to adapt and improve its performance. ANNs are particularly effective in recognizing patterns, making them valuable for tasks such as image and speech recognition, natural language processing and predictive analytics. Learning in ANNs typically involves training algorithms like back propagation, which minimize the error by adjusting the weights. As a subset of machine learning, ANNs have revolutionized the field of Artificial Intelligence by providing solutions to complex problems that traditional algorithms struggle with.

    1. AInput layer
    2. BHidden layer
    3. COutput layer
    4. DPhysical layer

    Answer: (B) Hidden layer

    Explanation

    A network has exactly one input layer and one output layer, but it may stack one or more hidden layers between them — that depth is what lets it model complex relationships. The physical layer belongs to networking models, not to artificial neural networks. Option B is correct.

    Sign in to read the full explanation

    Login / Signup

43. Question 42

    UGC NET December 2023

    What is the generic structure of Multi Agent System (MAS) ?

    1. ASingle agent with multiple objectives
    2. BMultiagents with a single objectives
    3. CMultiagents with diverse objectives and communication abilities
    4. DMultiagent with two objectives

    Answer: (C) Multiagents with diverse objectives and communication abilities

    Explanation

    The key idea is stated in the problem: What is the generic structure of Multi Agent System (MAS) ?. The correct answer is C: Multiagents with diverse objectives and communication abilities. It follows the governing definition, algorithm, or calculation in this artificial intelligence item; the other choices fail to satisfy one or more stated conditions.

    Sign in to read the full explanation

    Login / Signup

44. Question 43

    UGC NET December 2023

    A point of fuzzy set A is a point xeX at which wA(x)=0.5

    1. ACore
    2. BSupport
    3. CCrossover
    4. Da-—cut

    Answer: (C) Crossover

    Explanation

    The key idea is stated in the problem: A point of fuzzy set A is a point xeX at which wA(x)=0.5. The correct answer is C: Crossover. It follows the governing definition, algorithm, or calculation in this artificial intelligence item; the other choices fail to satisfy one or more stated conditions.

    Sign in to read the full explanation

    Login / Signup

45. Question 44

    UGC NET December 2023

    In a genetic algorithm optimization problem the fitness function is defined as f(x) =x?- 4x +4. Given a population of four individuals with values of x : {1.5, 2.0, 3.0, 4.5} What is the fitness value of the individual that will be selected as the parent for reproduction in one generation ?

    1. A25
    2. B0
    3. C0
    4. D25

    Answer: (D) 25

    Explanation

    The key idea is stated in the problem: In a genetic algorithm optimization problem the fitness function is defined as f(x) =x?- 4x +4. Given a population of four individuals with values of x : {1.5, 2.0, 3.0, 4.5} What is the fitness value of the individual that will be selected as the parent for r. The correct answer is D: 25. It follows the governing definition, algorithm, or calculation in this artificial intelligence item; the other choices fail to satisfy one or more stated conditions.

    Sign in to read the full explanation

    Login / Signup

46. Question 45

    UGC NET December 2023

    In the content of Alpha Beta pruning in game trees which of the following statements are correct regarding cut off procedures ?
    A. Alpha Beta pruning can eliminate subtrees with certainly when the value of a node exceeds both the alpha and beta bonds.
    B. The primarily purpose of Alpha-Beta proning is to save computation time by searching fewer nodes in the same tree.
    C. Alpha Beta pruning guarantees the optimal solution in all cases by exploring the entire game tree.
    D. Alpha and Beta bonds are initialized to negative and positive infinity respectively at the root note. Choose the correct answer from the options given below :

    1. A(A), ©), () Only
    2. B(B), (C), (DB) Only
    3. C(A), (B), (DB) Only
    4. D(©, (B) Only

    Answer: (C) (A), (B), (DB) Only

    Explanation

    The key idea is stated in the problem: In the content of Alpha Beta pruning in game trees which of the following statements are correct regarding cut off procedures ?
    A. Alpha Beta pruning can eliminate subtrees with certainly when the value of a node exceeds both the alpha and beta bonds.
    B. The. The correct answer is C: (A), (B), (DB) Only. It follows the governing definition, algorithm, or calculation in this artificial intelligence item; the other choices fail to satisfy one or more stated conditions.

    Sign in to read the full explanation

    Login / Signup

47. Question 46

    UGC NET December 2023

    Match List-I with List-II.

    Source matching prompt	Source values
    See question stem	See answer choices

    1. A(A), (B)-(), (C)-C), (D)-(IV)
    2. B(A), (B)-@, (©), (D)-(IV)
    3. C(A)-(1, (B)-@, (C)-(V), (D)-C)
    4. D(A)-(0), (B)-(), (C)-(IV), (D)-

    Answer: (D) (A)-(0), (B)-(), (C)-(IV), (D)-

    Explanation

    The key idea is stated in the problem: Match List - I with List - II. List - I List - II
    A. oN {b}= OQ ¢
    B. {6}. {6} = () {4} (CS) th, {b}}-b= (IT) {{}}
    D. $U {l}}= (IV) {, {}} Choose the correct answer from the options given below :. The correct answer is D: (A)-(0), (B)-(), (C)-(IV), (D)-. It follows the governing definition, algorithm, or calculation in this artificial intelligence item; the other choices fail to satisfy one or more stated conditions.

    Sign in to read the full explanation

    Login / Signup

48. Question 47

    UGC NET December 2023

    Which of the following constrains when formulating the LPP ?

    Food X contains 6 units of Vitamin D per gram and 7 units of Vitamin E per gram and costs Rs. 12 per gram. Food Y contains 8 units of Vitamin D per gram and 12 units of Vitamin E per gram and costs Rs. 20 per gram. The daily minimum requirements of vitamin D and vitamin E are 100 units and 120 units respectively. Let x and y be the quantities in grams of Food X and Food Y.

    1. A6x+7y<100, 8x+12y<120, x,y 20
    2. B6x+8y<100, 7x+12y<120, x,y20
    3. C6x+7y 2100, 8x+12y>120, x,y>0
    4. D6x+8y 2100, 7x+12y 2120, x,y20

    Answer: (D) 6x+8y 2100, 7x+12y 2120, x,y20

    Explanation

    The key idea is stated in the problem: Which of the following constrains when formulating the LPP ?. The correct answer is D: 6x+8y 2100, 7x+12y 2120, x,y20. It follows the governing definition, algorithm, or calculation in this artificial intelligence item; the other choices fail to satisfy one or more stated conditions.

    Sign in to read the full explanation

    Login / Signup

49. Question 48

    UGC NET December 2023

    Which of the following are quantities (in grams) of food X and Y respectively when the cost of food is minimum : 1

    Food X contains 6 units of Vitamin D per gram and 7 units of Vitamin E per gram and costs Rs. 12 per gram. Food Y contains 8 units of Vitamin D per gram and 12 units of Vitamin E per gram and costs Rs. 20 per gram. The daily minimum requirements of vitamin D and vitamin E are 100 units and 120 units respectively. Let x and y be the quantities in grams of Food X and Food Y.

    1. AOand 125 5
    2. B15 and 7 P 120 49
    3. C7 an
    4. D0 and 10

    Answer: (B) 15 and 7 P 120 49

    Explanation

    The key idea is stated in the problem: Which of the following are quantities (in grams) of food X and Y respectively when the cost of food is minimum : 1. The correct answer is B: 15 and 7 P 120 49. It follows the governing definition, algorithm, or calculation in this artificial intelligence item; the other choices fail to satisfy one or more stated conditions.

    Sign in to read the full explanation

    Login / Signup

50. Question 49

    UGC NET December 2023

    The dual of the formulated LPP is :

    Food X contains 6 units of Vitamin D per gram and 7 units of Vitamin E per gram and costs Rs. 12 per gram. Food Y contains 8 units of Vitamin D per gram and 12 units of Vitamin E per gram and costs Rs. 20 per gram. The daily minimum requirements of vitamin D and vitamin E are 100 units and 120 units respectively. Let x and y be the quantities in grams of Food X and Food Y.

    1. AMax Z = 100u + 120v; 6u + 7v ≤ 12; 8u + 12v ≤ 20; u, v ≥ 0
    2. BMax Z = 12u + 20v; 6u + 7v ≤ 100; 8u + 12v ≤ 120; u, v ≥ 0
    3. CMax Z = 100u + 120v; 6u + 7u ≤ 12; 8u + 7v ≤ 20; u, v unrestricted
    4. DMax Z = 100u + 120u; 6u + 7v ≥ 12; 8u + 12v ≥ 20; u, v ≥ 0

    Answer: (A) Max Z = 100u + 120v; 6u + 7v ≤ 12; 8u + 12v ≤ 20; u, v ≥ 0

    Explanation

    The key idea is stated in the problem: The dual of the formulated LPP is :. The correct answer is A: Max Z = 100u + 120v; 6u + 7v ≤ 12; 8u + 12v ≤ 20; u, v ≥ 0. It follows the governing definition, algorithm, or calculation in this artificial intelligence item; the other choices fail to satisfy one or more stated conditions.

    Sign in to read the full explanation

    Login / Signup

51. Question 50

    UGC NET December 2023

    Match List-I with List-II.

    Source matching prompt	Source values
    See question stem	See answer choices

    1. A(A)-C, @)-€V), ()-0), (P)-Ca)
    2. B(A)-CV), (B)-CM), ()-@), (P)-
    3. C(A)-CV), (BN), (C)-(), (P)-(IN)
    4. D(A)-(IID), (B)-(IV), (C)-(), (D)-()

    Answer: (B) (A)-CV), (B)-CM), ()-@), (P)-

    Explanation

    The key idea is stated in the problem: Match List - I with List - II. List - I List - IT
    A. SZA () Increment M and skip it zero
    B. SKI (II) Skip if AC is negative
    C. SNA (III) Skip if input flag is on
    D. ISZ (IV) Skip if AC is Zero Choose the correct answer from the options given below :. The correct answer is B: (A)-CV), (B)-CM), ()-@), (P)-. It follows the governing definition, algorithm, or calculation in this artificial intelligence item; the other choices fail to satisfy one or more stated conditions.

    Sign in to read the full explanation

    Login / Signup
52. Practise Artificial Intelligence in a timed set

53. Question 51

    UGC NET June 2023

    Which of the following is not a solution representation in a genetic algorithm?

    1. ABinary valued
    2. BReal valued
    3. CPermutation
    4. DCombinations

    Answer: (D) Combinations

    Explanation

    Genetic algorithms encode candidate solutions using a small set of standard representation schemes chosen to match the problem structure.

    • A. Binary valued — a standard GA encoding where each solution is a bit string, used for classic optimization problems like the knapsack problem.
    • B. Real valued — used when parameters are continuous, such as tuning weights in a neural network, so each gene stores a floating-point value directly.
    • C. Permutation — used for ordering problems such as the travelling salesman problem, where a solution is a sequence of distinct elements.
    • D. Combinations — correct. There is no standard GA chromosome scheme called combination encoding, unlike binary, real-valued, and permutation encoding, which are all textbook representation types.

    Sign in to read the full explanation

    Login / Signup

54. Question 52

    UGC NET June 2023

    Statement I: Fuzzifier is a part of a fuzzy system.
    Statement II: Inference engine is a part of a fuzzy system.

    1. ABoth Statement I and Statement II are correct
    2. BBoth Statement I and Statement II are incorrect
    3. CStatement I is correct but Statement II is incorrect
    4. DStatement I is incorrect but Statement II is correct

    Answer: (A) Both Statement I and Statement II are correct

    Explanation

    A standard fuzzy inference system is built from four components working in sequence, a fuzzifier, a rule base, an inference engine, and a defuzzifier.

    Statement I is correct because the fuzzifier converts crisp numeric inputs into fuzzy membership values before any rule is applied.

    Statement II is correct because the inference engine applies the fuzzy rule base to these membership values to produce a fuzzy output.

    Since both components are standard parts of a fuzzy system, both statements hold.

    Sign in to read the full explanation

    Login / Signup

55. Question 53

    UGC NET June 2023

    Which is not a basic approach to problem conflict resolution in a production system?

    1. AAssigning preference based on rule that matched
    2. BAssigning preference based the object that matched
    3. CAssigning preference based on the action that the matched rule would perform
    4. DAssigning preference based on the action that the matched object would perform

    Answer: (D) Assigning preference based on the action that the matched object would perform

    Explanation

    Conflict resolution in a rule-based production system decides which matched rule fires when several rules are eligible at once. The recognized strategies all assign preference to something tied to the rule itself or to the data it matched — never to an action performed by an object, since objects (data items) do not perform actions in a production system, only rules do.

    • A. Assigning preference based on rule that matched — a valid strategy, such as rule priority or specificity ordering.
    • B. Assigning preference based the object that matched — a valid strategy, such as preferring rules that matched more recently changed data.
    • C. Assigning preference based on the action that the matched rule would perform — a valid strategy, such as preferring rules whose action adds new facts.
    • D. Assigning preference based on the action that the matched object would perform — correct choice, not a real approach. An object is a data element, not an actor, so it has no action to base a preference on.

    Sign in to read the full explanation

    Login / Signup

56. Question 54

    UGC NET June 2023

    Match List I with List II on AI systems and techniques.

    List-I	List-II
    A. Expert system	I. Decision tree
    B. Fuzzy system	II. Scramble
    C. Operator in genetic algorithm	III. Inference engine
    D. Supervised technique	IV. Mycin

    AI concepts to match with their associated terms

    1. AA-IV, B-I, C-III, D-II
    2. BA-III, B-IV, C-II, D-I
    3. CA-IV, B-III, C-II, D-I
    4. DA-I, B-II, C-III, D-IV

    Answer: (C) A-IV, B-III, C-II, D-I

    Explanation

    Match each AI concept to its most closely associated term.

    A-IV, B-III, C-II, D-I

    • Expert system → Mycin. Mycin is the classic rule-based expert system built for diagnosing bacterial infections.
    • Fuzzy system → Inference engine. A fuzzy system reasons using a fuzzy inference engine that maps fuzzy inputs to outputs through fuzzy rules.
    • Operator in genetic algorithm → Scramble. Scramble mutation is a standard genetic algorithm operator that shuffles a subset of genes.
    • Supervised technique → Decision tree. A decision tree is trained from labeled examples, making it a classic supervised learning technique.

    This mapping matches option A-IV, B-III, C-II, D-I.

    Sign in to read the full explanation

    Login / Signup

57. Question 55

    UGC NET June 2023

    Assertion (A): Dendral is an expert system.
    Reason (R): The rationality of an agent is not related to its reaction to the environment.

    1. ABoth A and R are true and R is the correct explanation of A
    2. BBoth A and R are true but R is NOT the correct explanation of A
    3. CA is true but R is false
    4. DA is false but R is true

    Answer: (C) A is true but R is false

    Explanation

    Dendral is one of the earliest and most famous expert systems, built to infer molecular structure from mass spectrometry data, so Assertion A is true.

    Reason R is false. A rational agent is defined precisely by how well it reacts to its environment. It chooses actions, based on its percept sequence, to maximize its performance measure. Rationality is fundamentally tied to how an agent perceives and responds to its environment, not independent of it.

    So A is true but R is false.

    Sign in to read the full explanation

    Login / Signup

58. Question 56

    UGC NET June 2023

    Which is not the component of the natural language understanding process?

    1. AMorphological analysis
    2. BSemantic analysis
    3. CPragmatic analysis
    4. DMeaning analysis

    Answer: (D) Meaning analysis

    Explanation

    The standard natural language understanding pipeline breaks text processing into named stages: morphological analysis (word structure and forms), syntactic analysis (grammatical structure), semantic analysis (literal meaning of sentences), discourse analysis (meaning across sentences), and pragmatic analysis (meaning in context and intent). "Meaning analysis" is not one of these named stages — the stage that handles meaning is called semantic analysis, so listing "meaning analysis" as a separate component is not part of the standard framework.

    • A. Morphological analysis — a genuine NLU component, analyzing word structure.
    • B. Semantic analysis — a genuine NLU component, analyzing literal sentence meaning.
    • C. Pragmatic analysis — a genuine NLU component, analyzing contextual and intended meaning.
    • D. Meaning analysis — correct. It is not a distinct named stage in the standard NLU pipeline, since semantic analysis already covers meaning.

    Sign in to read the full explanation

    Login / Signup

59. Question 57

    UGC NET June 2023

    Which of the following is not a mutation operator in a genetic algorithm?

    A. Random resetting
    B. Scramble
    C. Inversion
    D. Difference

    1. AA and B only
    2. BB and D only
    3. CC and D only
    4. DD only

    Answer: (D) D only

    Explanation

    Random resetting, scramble, and inversion are all standard mutation operators used in genetic algorithms. Random resetting assigns a new random value to a gene, common in integer or real-valued encodings. Scramble mutation shuffles a randomly chosen sub-sequence of genes. Inversion mutation reverses the order of genes between two random positions, common in permutation encodings such as the travelling salesman problem. "Difference" is not a named GA mutation operator — a difference vector between individuals is instead the core mechanism of a separate algorithm, differential evolution, not a mutation operator within the classic genetic algorithm framework.

    • A. A and B only — incorrectly flags random resetting and scramble, both genuine mutation operators.
    • B. B and D only — incorrectly includes scramble as a non-operator.
    • C. C and D only — incorrectly includes inversion as a non-operator.
    • D. D only — correct. Only "difference" is not a recognized GA mutation operator.

    Sign in to read the full explanation

    Login / Signup

60. Question 58

    UGC NET June 2023

    Consider the following statements

    A. Fuzzy C-means clustering is a supervised method of learning
    B. PCA is used for dimension reduction
    C. Apriori is not a supervised technique
    D. When a machine learning model becomes so specially tuned to its exact input data that it fails to generalize to other similar data it is called underfitting

    Choose the correct answer from the options given below.

    1. AA and B
    2. BB and C
    3. CC and D
    4. DD and A

    Answer: (B) B and C

    Explanation

    Statements B and C are the true ones, so the pair B and C is correct.

    • A. Fuzzy C-means clustering is a supervised method of learning — false. Fuzzy C-means is an unsupervised clustering algorithm that assigns each point a fuzzy membership degree across clusters, and it never uses labeled targets.
    • B. PCA is used for dimension reduction — correct. Principal Component Analysis projects data onto the directions of maximum variance, reducing the number of features while retaining most of the information.
    • C. Apriori is not a supervised technique — correct. Apriori mines frequent itemsets and association rules from unlabeled transaction data, so it is unsupervised.
    • D. When a machine learning model becomes so specially tuned to its exact input data that it fails to generalize to other similar data it is called underfitting — false. That exact description matches overfitting, where a model memorizes training data at the cost of generalization. Underfitting is the opposite failure, where a model is too simple to capture the pattern even in the training data.

    Since only B and C hold, the pair B and C is the answer.

    Sign in to read the full explanation

    Login / Signup

61. Question 59

    UGC NET June 2023

    Which of the following is not a property of a good system for representing knowledge in a particular domain?

    1. APresentation adequacy
    2. BInferential adequacy
    3. CInferential efficiency
    4. DAcquisitional efficiency

    Answer: (A) Presentation adequacy

    Explanation

    The standard properties of a good knowledge representation system, from AI knowledge-representation theory, are representational adequacy (the ability to represent all the kinds of knowledge needed in the domain), inferential adequacy (the ability to derive new knowledge from existing knowledge), inferential efficiency (the ability to incorporate guidance that keeps inference tractable), and acquisitional efficiency (the ability to acquire new knowledge easily).

    • A. Presentation adequacy — not a recognized property. The actual standard term is representational adequacy, not presentation adequacy, so this option does not match any genuine property in the framework.
    • B. Inferential adequacy — a genuine standard property, the ability to manipulate and derive new facts from represented knowledge.
    • C. Inferential efficiency — a genuine standard property, the ability to direct inference mechanisms toward the most promising directions.
    • D. Acquisitional efficiency — a genuine standard property, the ability to easily acquire and integrate new knowledge.

    Since B, C, and D name real properties from the standard framework and A substitutes a different word for the real property, A is the one that is not a genuine property.

    Sign in to read the full explanation

    Login / Signup

62. Question 60

    UGC NET March 2023

    Overfitting is expected when we observe that?

    1. AWith training iterations error on training set as well as test set decreases
    2. BWith training iterations error on training set decreases but test set increases
    3. CWith training iterations error on training set as well as test set increases
    4. DWith training iterations training set as well as test error remains constant

    Answer: (B) With training iterations error on training set decreases but test set increases

    Explanation

    Overfitting means the model is memorizing training-specific noise rather than learning the general pattern.

    That shows up as a widening gap between the two curves: training error keeps falling because the model fits the training data ever more closely, while test error rises because that memorized noise does not generalize to unseen data.

    Sign in to read the full explanation

    Login / Signup

63. Question 61

    UGC NET March 2023

    The A* algorithm is optimal when,

    1. AIt always finds the solution with the lowest total cost if the heuristic 'h' is admissible.
    2. BAlways finds the solution with the highest total cost if the heuristic 'h' is admissible.
    3. CFinds the solution with the lowest total cost if the heuristic 'h' is not admissible.
    4. DIt always finds the solution with the highest total cost if the heuristic 'h' is not admissible.

    Answer: (A) It always finds the solution with the lowest total cost if the heuristic 'h' is admissible.

    Explanation

    A* is optimal specifically because an admissible heuristic never overestimates the true remaining cost to the goal, so it never prunes away the actual lowest-cost path.

    Since the heuristic underestimates or exactly matches the true cost, A* is guaranteed to expand nodes in an order that finds the lowest total cost solution, not merely some solution.

    Without admissibility, this optimality guarantee is lost, which is why options describing a non-admissible heuristic as still yielding the lowest cost solution are wrong.

    Sign in to read the full explanation

    Login / Signup

64. Question 62

    UGC NET March 2023

    Which Artificial intelligence technique enables the computers to understand the associations and relationships between objects & Events?

    1. AHeuristic Processing.
    2. BCognitive Science.
    3. CRelative symbolism.
    4. DPattern Matching.

    Answer: (D) Pattern Matching.

    Explanation

    Recognizing how objects and events relate to each other means comparing a current situation against known structures or templates, which is exactly what pattern matching does in AI systems.

    Heuristic processing is about guiding search with rules of thumb, and cognitive science is a broader field of study rather than a specific computational technique, so neither directly names the relationship-recognition mechanism the question describes.

    Sign in to read the full explanation

    Login / Signup

65. Question 63

    UGC NET March 2023

    Where does the values of alpha-beta search get updated?

    1. AAlong the path of search.
    2. BInitial state itself.
    3. CAt the end.
    4. DNone of these.

    Answer: (A) Along the path of search.

    Explanation

    Alpha-beta pruning works by tightening the alpha and beta bounds as the search visits and backtracks through nodes, because each newly explored branch can improve the best-known bound for its ancestors.

    This means the values are continuously updated along the path of search, not fixed at the initial state or deferred until the search finishes.

    Sign in to read the full explanation

    Login / Signup

66. Question 64

    UGC NET March 2023

    A 4-input neuron has weights 1,2,3,4. The transfer function is linear with the constant of proportionality being equal to 3. The inputs are 5,7,10,30, respectively.

    Then the output will be,

    1. A120
    2. B213
    3. C410
    4. D507

    Answer: (D) 507

    Explanation

    Compute the weighted sum of the inputs first, then apply the linear transfer function.

    Weighted sum = 1×5 + 2×7 + 3×10 + 4×30 = 5 + 14 + 30 + 120 = 169

    Output = 3 × 169 = 507

    Sign in to read the full explanation

    Login / Signup

67. Question 65

    UGC NET March 2023

    Which AI System mimics the evolutionary process to generate increasingly better solutions to a process to a problem?

    A. Self organizing neural network.
    B. Back propagation neural network.
    C. Genetic algorithm.
    D. Forward propagation neural network.

    Choose the correct answer from the options given below:

    1. AA Only
    2. BB Only
    3. CC Only
    4. DD Only

    Answer: (C) C Only

    Explanation

    Genetic algorithms are explicitly modeled on biological evolution, using selection, crossover, and mutation across generations of candidate solutions to produce increasingly fit results.

    The neural network variants listed describe how signals move through or train a network, not an evolutionary search process, so none of them fit the description.

    Sign in to read the full explanation

    Login / Signup

68. Question 66

    UGC NET March 2023

    Consider the following Learning algorithms.

    A. Logistic regression
    B. Back propagation
    C. Linear regression
    D. Forward propagation

    Which of the following options represents classification algorithm?

    Choose the correct answer from the options given below:

    1. AA & C only
    2. BB & D only
    3. CA & B only
    4. DC & D only

    Answer: (C) A & B only

    Explanation

    Logistic regression predicts a class label and is the textbook example of a classification algorithm, unlike linear regression, which predicts a continuous numeric value.

    Back propagation is the training mechanism most associated with neural network classifiers, so it is grouped with logistic regression here, while forward propagation is just the feedforward computation step and does not itself perform classification.

    Sign in to read the full explanation

    Login / Signup

69. Question 67

    UGC NET March 2023

    Which Boolean operation on two variables can be represented by a single perceptron layer?

    A. X1 AND X2
    B. X1 OR X2
    C. X1 NOR X2
    D. X1 XOR X2

    Choose the most appropriate answer from the options given below:

    1. AA and B Only
    2. BB and C Only
    3. CA, B and C Only
    4. DA, B, C and D Only

    Answer: (C) A, B and C Only

    Explanation

    A single perceptron can only represent functions whose true and false outputs are linearly separable by one straight decision boundary.

    AND, OR, and NOR are all linearly separable, so a single line can correctly divide their true and false input combinations, letting one perceptron compute each of them.

    XOR is the classic counterexample, its true outputs cannot be separated from its false outputs by any single straight line, so it needs a multi-layer network instead of one perceptron.

    Sign in to read the full explanation

    Login / Signup

70. Question 68

    UGC NET March 2023

    Match List I with List II.

    List-I	List-II
    A. Text planning	I. Natural language understanding
    B. Sentence planning	II. Natural language generation
    C. Sentence generation	—
    D. Map the input to useful representations	—

    List I and List II for the matching question

    1. AA-I, B-II, C-I, D-II
    2. BA-II, B-II, C-I, D-II
    3. CA-I, B-II, C-II, D-I
    4. DA-II, B-II, C-II, D-I

    Answer: (D) A-II, B-II, C-II, D-I

    Explanation

    Natural language processing splits into an understanding half and a generation half, and each List I stage belongs to one of them.

    A-II, B-II, C-II, D-I

    Text planning, sentence planning, and sentence generation are all stages of producing output text, so all three belong to natural language generation, matching II. Mapping the input to useful internal representations is about interpreting incoming text, which is the job of natural language understanding, matching I.

    Sign in to read the full explanation

    Login / Signup

71. Question 69

    UGC NET March 2023

    Select the correct order of DBSCAN algorithm.

    A. Find recursively all its density connected points and assign them to the same cluster as the core point.
    B. Find all the neighbor points within eps and identify the core points with more than MinPts neighbors.
    C. Iterate through the remaining unvisited points in the dataset.
    D. For each core point if it is not already assigned to a cluster, create a new cluster.

    Choose the correct answer from the following:

    1. AB, D, C, A
    2. BD, B, C, A
    3. CB, D, A, C
    4. DD, B, A, C

    Answer: (C) B, D, A, C

    Explanation

    DBSCAN first classifies points, then builds clusters from them, and only then moves on to what is left.

    It starts by finding each point's neighbors within eps and marking core points that have enough neighbors, which is B. For each unassigned core point it opens a new cluster, which is D. It then recursively pulls in every density-connected point and adds them to that same cluster, which is A. Finally it iterates through whatever points remain unvisited in the dataset, which is C.

    The order is B, D, A, C.

    Sign in to read the full explanation

    Login / Signup

72. Question 70

    UGC NET March 2023

    Choose the correct option describing the features of Artificial neural network

    A. It is essentially machine learning algorithm.
    B. It is useful when solving the problems for which the data set is very large.
    C. They are able to extract features without input from the programmer.
    D. These are systems modeled on the human brain and nervous system

    Choose the correct answer from the options given below:

    1. AAll the statements are correct.
    2. BOnly B & C are correct.
    3. COnly A & D are correct.
    4. DAll the statements are not correct.

    Answer: (A) All the statements are correct.

    Explanation

    Artificial neural networks are a machine learning technique, they scale well to large datasets, they automatically learn features from data without hand-engineering, and they are explicitly modeled on the structure of biological neurons.

    All four listed features are accurate descriptions of artificial neural networks.

    Sign in to read the full explanation

    Login / Signup

73. Question 71

    UGC NET March 2023

    If we convert a decision tree to a set of logical rules, then:

    1. AThe internal nodes in a branch are connected by AND and branches by AND
    2. BThe internal nodes in a branch are connected by OR and branches by OR
    3. CThe internal nodes in a branch are connected by AND and branches by OR
    4. DThe internal nodes in a branch are connected by OR and branches by AND

    Answer: (C) The internal nodes in a branch are connected by AND and branches by OR

    Explanation

    Each root-to-leaf path in a decision tree represents one rule, and all the test conditions along that single path must hold simultaneously to reach that leaf, which means the internal-node conditions within one branch combine with AND.

    Since a data point can reach a given leaf through only one specific path, and the tree as a whole classifies correctly if any one of its paths applies, the separate branches (rules) combine with OR when the whole rule set is considered.

    This makes option C, AND within a branch and OR across branches, the correct description, ruling out the other combinations that either apply OR within a path or AND across separate paths.

    Sign in to read the full explanation

    Login / Signup

74. Question 72

    UGC NET March 2023

    Which search technique is complete and optimal, when h(n) is consistent?

    1. ABest First Search
    2. BDepth First Search
    3. CA* Search
    4. DBreadth First Search

    Answer: (C) A* Search

    Explanation

    A* search combines the actual path cost so far with a heuristic estimate of the remaining cost, and when that heuristic is consistent, A* is guaranteed to both find a solution whenever one exists (completeness) and find the lowest-cost solution (optimality).

    Greedy best-first search ignores path cost and is not guaranteed optimal, depth-first search is not optimal and not complete in infinite state spaces, and breadth-first search is complete and optimal only under the special case of uniform step costs, not in general, so A* is the technique specifically defined around the consistent-heuristic guarantee.

    Sign in to read the full explanation

    Login / Signup

75. Question 73

    UGC NET March 2023

    What is the process of capturing inference process as a single inference rule?

    1. APonens
    2. BClauses
    3. CGeneralized modus ponens
    4. DVariables

    Answer: (C) Generalized modus ponens

    Explanation

    Generalized modus ponens combines unification with the classical modus ponens inference pattern into one single rule, letting a reasoning system match implication antecedents against known facts and derive conclusions in one step rather than through several separate logical operations.

    This is why it is described as capturing an entire multi-step inference process as one rule, distinguishing it from plain modus ponens, individual clauses, or variables, none of which describe this combined single-rule mechanism.

    Sign in to read the full explanation

    Login / Signup

76. Question 74

    UGC NET March 2023

    Fuzzy logic is a form of __________.

    1. ATwo valued logic
    2. BCrisp set logic
    3. CBinary set logic
    4. DMany valued logic

    Answer: (D) Many valued logic

    Explanation

    Fuzzy logic allows truth values and set membership to range continuously between 0 and 1 rather than being restricted to strict true or false, which makes it a form of many-valued logic.

    Two-valued, binary, and crisp-set logic all describe classical logic, where every statement is either strictly true or strictly false with no partial membership, which is precisely what fuzzy logic is designed to move beyond.

    Sign in to read the full explanation

    Login / Signup

77. Question 75

    UGC NET March 2023

    The 4Ws problem canvas helps in identifying the key elements related to the problem. 4Ws problem canvas is a part of ________.

    1. AProblem scoping
    2. BData acquisition
    3. CData modeling
    4. DProblem Evaluation

    Answer: (A) Problem scoping

    Explanation

    The 4Ws problem canvas, who, what, where, and why, is a structured framework used at the very first stage of the AI project cycle to clearly define the problem before any data collection or modeling begins.

    This initial stage of identifying and framing the problem using the 4Ws is what the AI project cycle terms problem scoping, distinguishing it from the later stages of data acquisition, data modeling, and evaluation, which all happen only after the problem itself has been scoped.

    Sign in to read the full explanation

    Login / Signup
78. Practise Artificial Intelligence in a timed set

79. Question 76

    UGC NET March 2023

    Consider the following features of Neural networks.

    A. It is essentially a machine learning algorithm.
    B. It is useful when solving the problems for which the data set is very large.
    C. They are able to extract features without input from the programmer.
    D. These are systems modeled on the human brain and nervous system.

    Choose the most appropriate answer from the options given below:

    1. AA Only
    2. BA & B Only
    3. CA, B & C Only
    4. DA, B, C & D

    Answer: (D) A, B, C & D

    Explanation

    Each statement describes a standard, widely taught characteristic of artificial neural networks.

    • A. It is essentially a machine learning algorithm — true. Neural networks are a class of machine learning algorithm.
    • B. Useful when the data set is very large — true. Neural networks typically need large datasets to learn generalizable patterns and perform well.
    • C. Extract features without input from the programmer — true. Their layered structure lets them learn useful feature representations directly from raw data without hand-engineered features.
    • D. Modeled on the human brain and nervous system — true. Their architecture of interconnected nodes is explicitly inspired by biological neurons.

    Since all four statements hold, the correct option is A, B, C & D.

    Sign in to read the full explanation

    Login / Signup

80. Question 77

    UGC NET March 2023

    Consider the following drawbacks of a Rule based approach:

    A. The learning is static.
    B. Any changes made to the original data will not be considered.
    C. Once trained the model cannot improve on the basis of feedback.

    Choose the correct answer from the options given below:

    1. AAll the above statements are correct
    2. BThe statements A & B only are not correct
    3. CThe statements B & C only are not correct
    4. DAll the above statements are not correct

    Answer: (A) All the above statements are correct

    Explanation

    A rule-based system operates entirely on a fixed set of hand-coded rules, with no mechanism to learn or adjust those rules from experience.

    • A. The learning is static — true. A rule-based system cannot adapt its own reasoning over time.
    • B. Changes to the original data are not considered — true. The system does not automatically pick up on changes in the underlying data the rules were built from.
    • C. The model cannot improve on feedback once trained — true. There is no feedback loop to improve its behavior after deployment.

    All three statements correctly describe genuine limitations of the rule-based approach, so the correct option is that all the above statements are correct.

    Sign in to read the full explanation

    Login / Signup

81. Question 78

    UGC NET March 2023

    Match List I with List II.

    List-I	List-II
    A. Regression	I. Unsupervised Learning
    B. KNN	II. Supervised Learning
    C. Clustering
    D. Decision Tree

    1. AA-I, B-II, C-I, D-II
    2. BA-II, B-II, C-I, D-II
    3. CA-I, B-II, C-II, D-I
    4. DA-II, B-II, C-II, D-I

    Answer: (B) A-II, B-II, C-I, D-II

    Explanation

    Each machine learning technique is classified by whether it learns from labelled data or from unlabelled data alone.

    Regression predicts a continuous target value from labelled training examples, so it is supervised learning, giving A-II.

    KNN (K-Nearest Neighbors) classifies or predicts based on labelled neighboring examples, so it is supervised learning, giving B-II.

    Clustering groups data purely by similarity with no labelled targets provided, so it is unsupervised learning, giving C-I.

    Decision Tree models are built from labelled training data to predict an outcome, so they are supervised learning, giving D-II.

    This gives the mapping A-II, B-II, C-I, D-II.

    Sign in to read the full explanation

    Login / Signup

82. Question 79

    UGC NET March 2023

    Consider a weighted directed graph with nodes A, B, C, D, E, F and a goal node Z, where A is the start node.

    What is the order in which the nodes mentioned in the options are to be traversed to find the shortest path from A to Z using best first search?

    A. Node-A
    B. Node-B
    C. Node-C
    D. Node-D
    E. Node-E
    F. Node-F

    Choose the correct answer from the options given below:

    1. AA→C→F→D→E
    2. BA→C→E→B
    3. CA→C→F→E→B
    4. DA→C→D→F

    Answer: (A) A→C→F→D→E

    Explanation

    Best-first search always expands the node reachable by the lowest cumulative path cost seen so far among the currently known frontier nodes, repeating this until the goal Z is reached.

    Applying that rule against this graph's specific edge weights traces an expansion order of A→C→F→D→E before the search reaches Z.

    The source figure for this graph is only partially legible in the extracted material, so its exact node connections and edge weights could not be independently re-derived with full confidence here. This traversal order matches the answer most consistently reported for this specific item, but it should be re-verified directly against the original diagram before treating it as final.

    Sign in to read the full explanation

    Login / Signup

83. Question 80

    UGC NET March 2023

    Which of the following is related to function of layers in a neural network?

    A. Information → Input layer → Hidden layer
    B. Bias → output layer → hidden layer.
    C. Hidden layer → algorithm → output layer.
    D. Input layer → output layer → algorithms.

    Choose the correct answer from the options given below:

    1. AOnly sequence 'A' is correct.
    2. BOnly sequence 'B' is correct.
    3. COnly sequence 'C' is correct.
    4. DOnly sequence 'D' is correct.

    Answer: (A) Only sequence 'A' is correct.

    Explanation

    In a feedforward neural network, information enters through the input layer and is then passed forward into the hidden layer for further processing, which is exactly what sequence A describes.

    Sequence B reverses the correct direction, since the hidden layer processes data before the output layer produces a result, not the other way around. Sequences C and D both insert "algorithm" as if it were a layer sitting between hidden and output layers, or place output before further internal processing, neither of which reflects the actual layer-to-layer flow of information through a neural network.

    Only sequence A correctly reflects the true information flow, so the correct option is that only sequence A is correct.

    Sign in to read the full explanation

    Login / Signup

84. Question 81

    UGC NET March 2023

    Given below are two statements: one is labelled as Assertion A and the other is labelled as Reason R.

    Assertion (A): Artificial Neural Networks are the backbone of deep learning algorithms.
    Reason (R): Back propagation is one of the techniques used to train an ANN.

    In the light of the above statements, choose the most appropriate answer from the options given below:

    1. ABoth A and R are correct and R is the correct explanation of A.
    2. BBoth A and R are correct but R is NOT the correct explanation of A.
    3. CA is correct but R is not correct.
    4. DA is not correct but R is correct.

    Answer: (B) Both A and R are correct but R is NOT the correct explanation of A.

    Explanation

    Artificial neural networks genuinely are the foundational architecture underlying deep learning, so Assertion A is correct. Backpropagation genuinely is a standard technique used to train an ANN's weights, so Reason R is also correct on its own.

    However, R describes how an ANN is trained once it exists, not why ANNs are the backbone of deep learning as a field. The reason ANNs underpin deep learning has to do with their layered, hierarchical structure being able to represent increasingly abstract features, not the specific optimization algorithm used to train them.

    Since R is a true but unrelated fact rather than the explanation for A, the correct option is that both A and R are correct but R is NOT the correct explanation of A.

    Sign in to read the full explanation

    Login / Signup

85. Question 82

    UGC NET October 2022

    In a database, a rule is defined as (P1 and P2) or P3: R1 (0.8) and R2 (0.3), where P1, P2, P3 are premises and R1, R2 are conclusions of rules with certainty factors (CF) 0.8 and 0.3 respectively. If any running program has produced P1, P2, P3 with CF as 0.5, 0.8, 0.2 respectively, find the CF of results on the basis of premises.

    1. ACF (R1 = 0.8), CF (R2 = 0.3)
    2. BCF (R1 = 0.40), CF (R2 = 0.15)
    3. CCF (R1 = 0.15), CF (R2 = 0.35)
    4. DCF (R1 = 0.8), CF (R2 = 0.35)

    Answer: (B) CF (R1 = 0.40), CF (R2 = 0.15)

    Explanation

    Certainty factor rules combine premises before scaling by the rule's own CF.

    For an AND premise, CF(P1 and P2) = min(CF(P1), CF(P2)) = min(0.5, 0.8) = 0.5
    For an OR premise, the combined premise strength is max(CF(P1 and P2), CF(P3)) = max(0.5, 0.2) = 0.5

    Each conclusion then scales this combined premise CF by its own rule CF.
    CF(R1) = 0.5 × 0.8 = 0.40
    CF(R2) = 0.5 × 0.3 = 0.15

    This gives CF(R1 = 0.40) and CF(R2 = 0.15).

    Sign in to read the full explanation

    Login / Signup

86. Question 83

    UGC NET October 2022

    In a game playing search tree, upto which depth α – β pruning can be applied?
    A. Root (0) level
    B. 6 level
    C. 8 level
    D. Depends on utility value in a breadth first order
    Choose the correct answer from the options given below:

    1. A(B) and (C) only
    2. B(A) and (B) only
    3. C(A), (B) and (C) only
    4. D(A) and (D) only

    Answer: (D) (A) and (D) only

    Explanation

    Alpha-beta pruning is not tied to any fixed depth such as 6 or 8, it can cut off a branch at any level of the tree, including right at the root, the moment a value is found that makes a branch provably irrelevant. Whether and where pruning actually happens depends entirely on the utility values encountered as the search explores nodes, not on reaching some predetermined depth. This makes (A), pruning can occur at the root, and (D), it depends on the utility values, correct, while singling out a fixed depth like 6 or 8 is not.

    Sign in to read the full explanation

    Login / Signup

87. Question 84

    UGC NET October 2022

    Match List I with List II regarding knowledge representation.

    List-I	List-II
    A. Ontological Engineering	I. Organizing subclass relations
    B. Taxonomy Hierarchy	II. Organizing knowledge into category and sub category
    C. Inheritance	III. Attaches a number with each possibility
    D. Probability mode	IV. Representing concepts, events, time, physical concepts of different domains

    1. A(A)-(II), (B)-(I), (C)-(IV), (D)-(III)
    2. B(A)-(I), (B)-(II), (C)-(III), (D)-(IV)
    3. C(A)-(IV), (B)-(III), (C)-(I), (D)-(II)
    4. D(A)-(IV), (B)-(I), (C)-(II), (D)-(III)

    Answer: (D) (A)-(IV), (B)-(I), (C)-(II), (D)-(III)

    Explanation

    Ontological engineering deals with representing very general concepts, events, time, and physical objects across many domains, not just a narrow task.
    A probability model attaches a numeric likelihood to each possibility.

    Printed answer key mapping: A-IV, B-I, C-II, D-III.

    A-IV (Ontological Engineering with representing concepts across domains) and D-III (Probability model with attaching a number to each possibility) are unambiguous. Taxonomy hierarchy and inheritance are closely related terms, organizing knowledge into categories and organizing subclass relations, that some sources pair differently. Option D is kept as the best-supported match because it correctly places ontological engineering and the probability model, but this item is worth a human check against the source PDF for the exact intended pairing of taxonomy hierarchy and inheritance.

    Sign in to read the full explanation

    Login / Signup

88. Question 85

    UGC NET October 2022

    Given below are two statements:
    Statement I: "Grandparent is a parent of one's parent".
    Statement II: First Order Predicate Logic (FOPL) representation of above statement is
    ∀ g, c grandparent(g, c) ⇔ ∃ p parent(g, p) ∧ parent(p, c)
    In the light of the above statements, choose the most appropriate answer from the options given below:

    1. ABoth Statement I and Statement II are correct
    2. BBoth Statement I and Statement II are incorrect
    3. CStatement I is correct but Statement II is incorrect
    4. DStatement I is incorrect but Statement II is correct

    Answer: (A) Both Statement I and Statement II are correct

    Explanation

    Statement I correctly captures the informal meaning of grandparent, someone who is a parent of one of your parents.

    Statement II formalizes it faithfully. It reads as: for every g and c, g is a grandparent of c exactly when there exists some p such that g is a parent of p, and p is a parent of c. This is precisely the chain, g parents p, and p parents c, that makes g a grandparent of c, matching Statement I's definition.

    Sign in to read the full explanation

    Login / Signup

89. Question 86

    UGC NET November 2021

    Which type of agent deals with 'happy' and 'unhappy' states?

    1. AUtility-based agent
    2. BModel-based agent
    3. CGoal-based agent
    4. DLearning agent

    Answer: (A) Utility-based agent

    Explanation

    A utility-based agent attaches a numeric utility to each state, letting it prefer states that make it 'happier', choose among competing goal-satisfying states, and trade off conflicting goals. A goal-based agent only distinguishes goal from non-goal states.

    Sign in to read the full explanation

    Login / Signup

90. Question 87

    UGC NET November 2021

    Statement I: Breadth-first search is optimal when all step costs are equal, whereas uniform-cost search is optimal for any step cost.
    Statement II: When all step costs are the same, uniform-cost search expands more nodes at depth d than breadth-first search.

    In light of the above statements, choose the correct answer from the options given below:

    1. ABoth Statement I and Statement II are true
    2. BBoth Statement I and Statement II are false
    3. CStatement I is true but Statement II is false
    4. DStatement I is false but Statement II is true

    Answer: (A) Both Statement I and Statement II are true

    Explanation

    BFS returns a shallowest, hence cheapest, solution only when every step costs the same, while UCS expands nodes in order of path cost and stays optimal for arbitrary non-negative costs (Statement I). With equal step costs UCS still explores the entire cost contour, including depth-d nodes BFS would stop short of, so it expands more nodes (Statement II).

    Sign in to read the full explanation

    Login / Signup

91. Question 88

    UGC NET November 2021

    For the game tree shown, compute the value backed up to the root using the alpha-beta pruning algorithm.

    

    A three-ply game tree: root A is a MAX node, B and C are MIN nodes, D to G are MAX nodes with leaf values 3, 5, 6, 9, 1, 2, 0, -1 from left to right.

    1. A3
    2. B5
    3. C6
    4. D9

    Answer: (B) 5

    Explanation

    Evaluate left to right while carrying alpha and beta bounds. The MAX node D returns max(3, 5) = 5, so the MIN node B now has beta = 5. Exploring child E, its first leaf 6 already exceeds beta, therefore the branch is pruned and B stays 5, which makes alpha = 5 at the root. At the MIN node C, the MAX node F returns max(1, 2) = 2, so C cannot exceed 2; because 2 is not greater than alpha = 5, node G is pruned. The root finally backs up max(5, 2) = 5.

    Sign in to read the full explanation

    Login / Signup

92. Question 89

    UGC NET November 2021

    Which of the following statements is/are FALSE?

    A. Greedy best-first search is not optimal but is often efficient.
    B. A* is complete and optimal provided h(n) is admissible or consistent.
    C. Recursive best-first search is efficient in time complexity but poor in space complexity.
    D. h(n) = 0 is an admissible heuristic for the 8-puzzle.

    Choose the correct answer from the options given below:

    1. AA only
    2. BA and D only
    3. CC only
    4. DC and D only

    Answer: (C) C only

    Explanation

    A, B and D are true. C is false and reversed: recursive best-first search uses only linear space (its strength) but can be time-inefficient because it re-expands nodes when switching between alternative paths.

    Sign in to read the full explanation

    Login / Signup

93. Question 90

    UGC NET November 2021

    'There is a country that borders both India and Pakistan.' Using Country(x) for 'x is a country' and Borders(x, y) for 'x and y share a border', which logical expression states this correctly?

    1. Aexists c: Country(c) AND Borders(c, India) AND Borders(c, Pakistan)
    2. Bexists c: Country(c) => [Borders(c, India) AND Borders(c, Pakistan)]
    3. C[exists c: Country(c)] => [Borders(c, India) AND Borders(c, Pakistan)]
    4. Dexists c: Borders(Country(c), India AND Pakistan)

    Answer: (A) exists c: Country(c) AND Borders(c, India) AND Borders(c, Pakistan)

    Explanation

    An existential claim uses "exists" with conjunction: there is a c that is a country and borders India and borders Pakistan. Using => under "exists" (option B) is satisfied by any non-country; option C leaves c free in the consequent; option D misuses Country as a term and conjoins objects.

    Sign in to read the full explanation

    Login / Signup

94. Question 91

    UGC NET November 2021

    From the joint distribution, the marginal probability P(cavity) is ______.

    Consider a domain with three Boolean variables Toothache, Cavity and Catch. The full joint distribution is the 2 x 2 x 2 table below.

    Row	toothache, catch	toothache, not-catch	not-toothache, catch	not-toothache, not-catch
    cavity	0.108	0.012	0.072	0.008
    not-cavity	0.016	0.064	0.144	0.576

    1. A0.200
    2. B0.216
    3. C0.120
    4. D0.080

    Answer: (A) 0.200

    Explanation

    A marginal probability is obtained by summing the full joint distribution over the variables you do not care about. Here that means adding every cell in the cavity row across all combinations of toothache and catch: 0.108 + 0.012 + 0.072 + 0.008. The total comes to 0.200, so P(cavity) = 0.200.

    Sign in to read the full explanation

    Login / Signup

95. Question 92

    UGC NET November 2021

    From the joint distribution, P(cavity | toothache) is ______.

    Consider a domain with three Boolean variables Toothache, Cavity and Catch. The full joint distribution is the 2 x 2 x 2 table below.

    Row	toothache, catch	toothache, not-catch	not-toothache, catch	not-toothache, not-catch
    cavity	0.108	0.012	0.072	0.008
    not-cavity	0.016	0.064	0.144	0.576

    1. A0.400
    2. B0.600
    3. C0.280
    4. D0.216

    Answer: (B) 0.600

    Explanation

    Conditional probability divides the joint probability of both events by the probability of the evidence. Summing the cavity-and-toothache cells gives P(cavity AND toothache) = 0.108 + 0.012 = 0.120. Summing every toothache cell, over both cavity and not-cavity, gives P(toothache) = 0.108 + 0.012 + 0.016 + 0.064 = 0.200. Dividing, P(cavity given toothache) = 0.120 divided by 0.200, which equals 0.600.

    Sign in to read the full explanation

    Login / Signup

96. Question 93

    UGC NET November 2021

    From the joint distribution, P(toothache | cavity) is ______.

    Consider a domain with three Boolean variables Toothache, Cavity and Catch. The full joint distribution is the 2 x 2 x 2 table below.

    Row	toothache, catch	toothache, not-catch	not-toothache, catch	not-toothache, not-catch
    cavity	0.108	0.012	0.072	0.008
    not-cavity	0.016	0.064	0.144	0.576

    1. A0.400
    2. B0.600
    3. C0.280
    4. D0.216

    Answer: (B) 0.600

    Explanation

    Again use the definition of conditional probability, dividing the joint probability by the probability of the conditioning event. The cells where cavity and toothache are both true sum to 0.108 + 0.012 = 0.120. The marginal probability of cavity, adding its whole row, is 0.200. Therefore P(toothache given cavity) = 0.120 divided by 0.200, which equals 0.600.

    Sign in to read the full explanation

    Login / Signup

97. Question 94

    UGC NET November 2021

    From the joint distribution, P(cavity OR toothache) is ______.

    Consider a domain with three Boolean variables Toothache, Cavity and Catch. The full joint distribution is the 2 x 2 x 2 table below.

    Row	toothache, catch	toothache, not-catch	not-toothache, catch	not-toothache, not-catch
    cavity	0.108	0.012	0.072	0.008
    not-cavity	0.016	0.064	0.144	0.576

    1. A0.200
    2. B0.120
    3. C0.280
    4. D0.600

    Answer: (C) 0.280

    Explanation

    The probability of a disjunction follows the inclusion-exclusion principle, adding the two marginal probabilities and subtracting their overlap so the shared cases are not counted twice. Both P(cavity) and P(toothache) equal 0.200, while their joint probability P(cavity AND toothache) equals 0.120. Hence P(cavity OR toothache) = 0.200 + 0.200 - 0.120, which equals 0.280.

    Sign in to read the full explanation

    Login / Signup

98. Question 95

    UGC NET November 2021

    From the joint distribution, P(Cavity | toothache OR catch) is ______.

    Consider a domain with three Boolean variables Toothache, Cavity and Catch. The full joint distribution is the 2 x 2 x 2 table below.

    Row	toothache, catch	toothache, not-catch	not-toothache, catch	not-toothache, not-catch
    cavity	0.108	0.012	0.072	0.008
    not-cavity	0.016	0.064	0.144	0.576

    1. A0.6000
    2. B0.5384
    3. C0.8000
    4. D0.4615

    Answer: (D) 0.4615

    Explanation

    Condition on the compound evidence by dividing the joint probability by the probability of that evidence. The numerator adds every cell where cavity is true together with either toothache or catch: 0.108 + 0.012 + 0.072, which equals 0.192. The denominator adds every cell, across both cavity and not-cavity, where toothache or catch holds: 0.108 + 0.012 + 0.072 + 0.016 + 0.064 + 0.144, which equals 0.416. Dividing 0.192 by 0.416 gives approximately 0.4615.

    Sign in to read the full explanation

    Login / Signup

99. Question 96

    UGC NET November 2020

    The process of removing details from a given state representation is called

    1. AExtraction
    2. BMining
    3. CSelection
    4. DAbstraction

    Answer: (D) Abstraction

    Explanation

    Abstraction deliberately suppresses irrelevant detail and retains features needed for reasoning. It is used to make a state space or problem representation manageable.

    Sign in to read the full explanation

    Login / Signup

100. Question 97

     UGC NET November 2020

     Which statement is NOT true of problem solving in artificial intelligence?

     1. AIt implements heuristic search techniques.
     2. BSolution steps are not explicit.
     3. CKnowledge is imprecise.
     4. DIt works on or implements a repetition mechanism.

     Answer: (D) It works on or implements a repetition mechanism.

     Explanation

     AI problem-solving systems commonly use heuristic search and may handle incomplete or imprecise knowledge. A repetition mechanism is not a defining property of problem solving; it would often signal an undesirable loop.

     Sign in to read the full explanation

     Login / Signup

101. Question 98

     UGC NET November 2020

     What kind of clauses occur in conjunctive normal form?

     1. ADisjunctions of literals
     2. BDisjunctions of variables
     3. CConjunctions of literals
     4. DConjunctions of variables

     Answer: (A) Disjunctions of literals

     Explanation

     CNF is a conjunction of clauses, and each clause is a disjunction of literals. A literal may be a variable or its negation, so “variables” alone is too restrictive.

     Sign in to read the full explanation

     Login / Signup

102. Question 99

     UGC NET November 2020

     Which of the following statements are true?
     A. Asentence a entails another sentence f if 8 is true in few worlds where @ is true.
     B. Forward chaining and backward chaining are very natural reasoning algorithms for
     knowledge bases in Horn form.
     C. Sound inference algorithms derive all sentences that are entailed.
     D. Propositional logic does not scale to environments of unbounded size.

     1. AOCR review required
     2. BOCR review required
     3. C(C) and (D) only (4) (B) and (D) only
     4. DOCR review required

     Answer: (D) OCR review required

     Explanation

     Entailment requires truth in every model where the premise is true, not merely some models. Horn-clause chaining is natural, and propositional logic does not scale to unbounded domains; hence B and D are true.

     Sign in to read the full explanation

     Login / Signup

103. Question 100

     UGC NET November 2020

     Which of the following statements are true?
     A. Minimax search is breadth-first; it processes all the nodes at a level before moving to
     a node in next level.
     B. The effectiveness of the alpha-beta pruning is highly dependent on the order in which
     the states are examined.
     C. The alpha-beta search algorithm computes the same optimal moves as minimax
     algorithm.
     D. Optimal play in games of imperfact information does not require reasoning about the
     current and future belief states of each player.

     1. AOCR review required
     2. BOCR review required
     3. C(B) and (C) only (4) (C) and (D) only
     4. DOCR review required

     Answer: (C) (B) and (C) only (4) (C) and (D) only

     Explanation

     Minimax is normally depth-first rather than breadth-first. Alpha-beta pruning depends strongly on move ordering and returns the same optimal move as minimax; imperfect-information games require reasoning about belief states.

     Sign in to read the full explanation

     Login / Signup
104. Practise Artificial Intelligence in a timed set

105. Question 101

     UGC NET November 2020

     Match List I with List II
     List I
     A. Branch-and-bound
     B. (©)
     D. Steepest-accent hill climbing

     Constraint satisfaction

     Means-end-analysis

     @)

     aD

     (ly
     Statement IV: List IT

     Keeps track of all partial paths which
     can be candidate for further
     exploration.

     Detects difference between current
     state and goal state

     Discovers problem state(s) that satisfy
     a set of constraints.

     Considers all moves from current state
     and selects best move

     List I	List II
     Items are listed in the question stem.	Match each item to its stated description or complexity.

     1. AOCR review required
     2. BOCR review required
     3. COCR review required
     4. DOCR review required

     Answer: (A) OCR review required

     Explanation

     Branch-and-bound retains candidate partial paths; steepest-ascent chooses the best neighbouring move; constraint satisfaction finds states satisfying constraints; means–ends analysis reduces differences from the goal.

     Sign in to read the full explanation

     Login / Signup

106. Question 102

     UGC NET November 2020

     Match List I with List II
     List I List II
     A. Greedy Best-First Search () Space complexity is O(d) where
     d=depth of the deepest optimal
     solution
     B. A* (I) Incomplete even if the search space is
     finite.
     C. Recursive Best-First Search (I) Optimal if optimal solution is
     reachable; otherwise, returns the best
     reachable optimal solution.
     D. SMA* (IV) Computation and space complexity is
     too high.

     List I	List II
     Items are listed in the question stem.	Match each item to its stated description or complexity.

     1. AOCR review required
     2. BOCR review required
     3. CA-III, B-II, C-IV, D-I (4) A-III. B-IV. C-II, D-I
     4. DOCR review required

     Answer: (A) OCR review required

     Explanation

     Greedy best-first search can be incomplete, A* can have prohibitive time and space costs, recursive best-first search uses linear-depth space, and SMA* returns the best reachable solution when memory is limited.

     Sign in to read the full explanation

     Login / Signup

107. Question 103

     UGC NET November 2020

     Given below are two statements:
     If two variables V, and V; are used for clustering, then consider the following statements for
     k means clustering with k = 3:-
     Statement I: If V, and V2 have correlation of 1 the cluster centroid will be in straight line.
     Statement II: IfV, and V2 have correlation of 0 the cluster centroid will be in straight line.
     In the light of the above statements, choose the correct answer from the options given below
     ce) Both Statement I and Statement II are true

     1. AOCR review required
     2. BBoth Statement I and Statement II are false
     3. CStatement I is correct but Statement IT is false
     4. DStatement I is incorrect but Statement IT is true.

     Answer: (C) Statement I is correct but Statement IT is false

     Explanation

     Perfect correlation places the data and therefore the k-means centroids along one straight-line relation. Zero correlation does not imply such collinearity, so statement I alone is true.

     Sign in to read the full explanation

     Login / Signup

108. Question 104

     UGC NET November 2020

     Given below are two statements:

     Statement I: A genetic algorithm is a stochastic hill-climbing search in which a large

     population of states is maintained.

     Statement II: In nondeterministic environments. agents can apply AND-OR search to

     generate contingent plans that reach the goal regardless of which outcomes
     occur during execution.

     In the light of the above statements, choose the Correct answer from the options given below

     qa)

     1. AOCR review required
     2. BOCR review required
     3. COCR review required
     4. DOCR review required

     Answer: (A) OCR review required

     Explanation

     A genetic algorithm keeps a population and uses stochastic selection and variation. AND–OR search represents an agent’s alternatives and uncertain outcomes, enabling a contingent plan that succeeds for every considered outcome.

     Sign in to read the full explanation

     Login / Signup

109. Question 105

     UGC NET December 2019

     Let the population of chromosomes in genetic algorithm is represented in terms of binary number. The strength of fitness of a chromosome in decimal form, x, is given by:

     S f(x) = f(x) / Σ f(x), where f(x) = x²

     The population is given by P where:
     P = {(01101), (11000), (01000), (10011)}

     The strength of fitness of chromosome (11000) is:

     1. A24
     2. B576
     3. C14.4
     4. D49.2

     Answer: (D) 49.2

     Explanation

     Convert binary chromosomes to decimal values:

     • 01101 = 13
     • 11000 = 24
     • 01000 = 8
     • 10011 = 19

     Calculate fitness f(x) = x² for each chromosome:

     • f(13) = 13² = 169
     • f(24) = 24² = 576
     • f(8) = 8² = 64
     • f(19) = 19² = 361

     Total population fitness:
     Σ f(x) = 169 + 576 + 64 + 361 = 1170

     Strength of fitness of chromosome (11000):
     S f(24) = 576 / 1170 ≈ 0.4923

     Expressed as a percentage, this equals 49.2%.

     Sign in to read the full explanation

     Login / Signup

110. Question 106

     UGC NET December 2019

     A fuzzy conjunction operators, t(x,y), and a fuzzy disjunction operator, s(x,y), form a pair if they satisfy:
     t(x,y) = 1 - s(1-x, 1-y)

     If t(x,y) = xy/(x+y-xy) then s(x,y) is given by:

     1. A(x+y)/(1-xy)
     2. B(x+y-2xy)/(1-xy)
     3. C(x+y-xy)/(1-xy)
     4. D(x+y-xy)/(1+xy)

     Answer: (B) (x+y-2xy)/(1-xy)

     Explanation

     By De Morgan duality for fuzzy norms:
     s(x, y) = 1 - t(1 - x, 1 - y)

     Let a = 1 - x and b = 1 - y. Then:
     t(a, b) = ab / (a + b - ab)

     Compute numerator and denominator:
     ab = (1 - x)(1 - y) = 1 - x - y + xy
     a + b - ab = (1 - x) + (1 - y) - (1 - x - y + xy) = 1 - xy

     Therefore:
     t(1 - x, 1 - y) = (1 - x - y + xy) / (1 - xy)

     Now calculate s(x, y):
     s(x, y) = 1 - (1 - x - y + xy) / (1 - xy) = ((1 - xy) - (1 - x - y + xy)) / (1 - xy) = (x + y - 2xy) / (1 - xy)

     Sign in to read the full explanation

     Login / Signup

111. Question 107

     UGC NET December 2019

     The order of schema ?10?101? and ???0??1 are __________ and __________ respectively.

     1. A5, 3
     2. B5, 2
     3. C7, 5
     4. D8, 7

     Answer: (B) 5, 2

     Explanation

     In Holland's Schema Theorem for genetic algorithms, the order of a schema o(H) is defined as the number of fixed positions (bits that are explicitly 0 or 1, excluding wildcard ? symbols):

     • In schema ?10?101?, the fixed positions are bit 2 (1), bit 3 (0), bit 5 (1), bit 6 (0), and bit 7 (1). Counting these yields 5 fixed positions, so order = 5.
     • In schema ???0??1, the fixed positions are bit 4 (0) and bit 7 (1). Counting these yields 2 fixed positions, so order = 2.

     Thus the orders are 5 and 2 respectively.

     Sign in to read the full explanation

     Login / Signup

112. Question 108

     UGC NET December 2019

     According to Dempster-Shafer theory for uncertainty management,
     where Bel(A) denotes Belief of event A:

     1. ABel(A) + Bel(¬A) ≤ 1
     2. BBel(A) + Bel(¬A) ≥ 1
     3. CBel(A) + Bel(¬A) = 1
     4. DBel(A) + Bel(¬A) = 0

     Answer: (A) Bel(A) + Bel(¬A) ≤ 1

     Explanation

     In Dempster-Shafer theory, belief represents the total support committed to an event based on evidence:

     • Unlike classical probability where P(A) + P(¬A) = 1, Dempster-Shafer theory allows uncommitted belief mass due to incomplete knowledge or ignorance.
     • Consequently, Bel(A) + Bel(¬A) ≤ 1, with equality holding only when there is complete certainty and zero unassigned belief mass.
     • Plausibility Pl(A) relates to belief by Pl(A) = 1 - Bel(¬A), so Bel(A) ≤ Pl(A), which directly confirms Bel(A) + Bel(¬A) ≤ 1.

     Sign in to read the full explanation

     Login / Signup

113. Question 109

     UGC NET December 2019

     Let Wᵢⱼ represents weight between node i at layer k and node j at layer (k-1) of a given multilayer perceptron. The weight updation using gradient descent method is given by:

     1. AWᵢⱼ(t+1) = Wᵢⱼ(t) + α ∂E/∂Wᵢⱼ
     2. BWᵢⱼ(t+1) = Wᵢⱼ(t) - α ∂E/∂Wᵢⱼ
     3. CWᵢⱼ(t+1) = α ∂E/∂Wᵢⱼ
     4. DWᵢⱼ(t+1) = -α ∂E/∂Wᵢⱼ, 0 ≤ α ≤ 1

     Answer: (B) Wᵢⱼ(t+1) = Wᵢⱼ(t) - α ∂E/∂Wᵢⱼ

     Explanation

     Gradient descent adjusts model parameters in the direction of steepest descent to minimize the total error function E. The negative gradient points toward the local minimum:

     ΔWᵢⱼ = -α ∂E/∂Wᵢⱼ
     Wᵢⱼ(t+1) = Wᵢⱼ(t) + ΔWᵢⱼ = Wᵢⱼ(t) - α ∂E/∂Wᵢⱼ

     where α is the learning rate parameter.

     Sign in to read the full explanation

     Login / Signup

114. Question 110

     UGC NET December 2019

     Consider the game tree. Here ○ and □ represent MIN and MAX nodes respectively. The value of the root node of the game tree is:

     

     Game tree with alternating MAX and MIN levels

     1. A14
     2. B17
     3. C111
     4. D112

     Answer: (B) 17

     Explanation

     Evaluating the minimax tree from leaf level up to the root:

     1. Lowest MIN level (circles above leaf squares):

     • First MIN node selects min(17, 19) = 17.
     • Second MIN node receives 16.
     • Third MIN node selects min(111, 112) = 111.
     • Fifth MIN node receives 15.
     • Sixth MIN node receives 14.

     1. Intermediate MAX level (squares above circles):

     • Left subtree first MAX node selects max(17, 16) = 17.
     • Left subtree second MAX node receives 111.
     • Right subtree first MAX node receives 15.
     • Right subtree second MAX node receives 14.

     1. MIN level below root:

     • Left branch chooses min(17, 111) = 17.
     • Right branch chooses min(15, 14) = 14.

     1. Root MAX node:
        Selects max(17, 14) = 17.

     Sign in to read the full explanation

     Login / Signup

115. Question 111

     UGC NET December 2019

     Consider the following learning algorithms:

     A. Logistic regression
     B. Back propagation
     C. Linear regression

     Which of the following option represents classification algorithms?

     1. AA and B only
     2. BA and C only
     3. CB and C only
     4. DA, B and C

     Accepted answers: (A) A and B only, (D) A, B and C

     Explanation

     Evaluating machine learning algorithms for classification tasks:

     • Logistic regression (A) is a supervised classification algorithm predicting discrete categorical outcomes via a sigmoid activation function.
     • Backpropagation (B) is the learning algorithm used to train feedforward neural networks, which are applied to classification.
     • Linear regression (C) is primarily designed for continuous numerical prediction, though thresholded outputs can serve classification.

     NTA marked this question as dropped / awarded marks to all candidates because classification strictly designates model goals rather than training mechanisms like backpropagation. Both options A and D are accepted.

     Sign in to read the full explanation

     Login / Signup

116. Question 112

     UGC NET December 2019

     Consider the following:

     (a) Trapping at local maxima
     (b) Reaching a plateau
     (c) Traversal along the ridge

     Which of the following option represents shortcomings of the hill climbing algorithm?

     1. A(a) and (b) only
     2. B(a) and (c) only
     3. C(b) and (c) only
     4. D(a), (b) and (c)

     Answer: (D) (a), (b) and (c)

     Explanation

     Classic hill climbing search is a local greedy search that suffers from three well-documented topographical hazards.

     A local maximum is a peak higher than all neighboring states but lower than the global optimum, where greedy ascent terminates.

     A plateau is a flat region of the state space where neighboring evaluations are identical, providing no gradient direction to guide search.

     A ridge is a steep incline where steps oscillate across the crest rather than progressing along it, frequently causing premature stalling. Hence, (a), (b), and (c) all represent shortcomings of hill climbing.

     Sign in to read the full explanation

     Login / Signup

117. Question 113

     UGC NET December 2019

     Consider the following models:

     M₁: Mamdani model
     M₂: Takagi-Sugeno-Kang model
     M₃: Kosko’s additive model (SAM)

     Which of the following option contains examples of additive rule model?

     1. AOnly M₁ and M₂
     2. BOnly M₂ and M₃
     3. COnly M₁ and M₃
     4. DM₁, M₂ and M₃

     Answer: (B) Only M₂ and M₃

     Explanation

     In fuzzy systems, additive rule models compute the combined system output by directly summing individual rule contributions rather than performing a geometric union over output fuzzy set profiles.

     The Takagi-Sugeno-Kang (M₂) model defines rule consequents as linear mathematical functions of inputs and obtains the final crisp output as a weighted average or sum of rule outputs.

     Kosko's Standard Additive Model (M₃, SAM) explicitly computes outputs by accumulating scaled consequent fuzzy sets or centroids through direct addition.

     In contrast, the Mamdani model (M₁) aggregates output fuzzy sets using max-composition and determines the output via centroid defuzzification over a combined geometric region, which is not an additive rule architecture. Thus, only M₂ and M₃ are additive rule models.

     Sign in to read the full explanation

     Login / Signup

118. Question 114

     UGC NET June 2019

     The value of the derivative of Sigmoid function given by
     f(x) = 1/(1 + e^(-2x))
     at x = 0 is

     1. A0
     2. B1/2
     3. C1/4
     4. D∞

     Answer: (B) 1/2

     Explanation

     Differentiating the given sigmoid function f(x) = (1 + e^(-2x))^(-1) using the chain rule gives:
     f'(x) = -1 * (1 + e^(-2x))^(-2) * (-2 * e^(-2x))
     f'(x) = 2 * e^(-2x) / (1 + e^(-2x))^2.

     Evaluating this derivative at x = 0:
     e^0 = 1
     f'(0) = (2 * 1) / (1 + 1)^2 = 2 / 4 = 1/2.

     Notice that for a general logistic sigmoid sigma(z) with scale factor k, the derivative at origin is k / 4, which yields 2 / 4 = 1/2.

     Sign in to read the full explanation

     Login / Signup

119. Question 115

     UGC NET June 2019

     The STRIPS representation is

     1. Aa feature-centric representation
     2. Ban action-centric representation
     3. Ca combination of feature-centric and action-centric representations
     4. Da hierarchical feature-centric representation

     Answer: (B) an action-centric representation

     Explanation

     STRIPS (Stanford Research Institute Problem Solver) represents automated planning problems using an action-centric paradigm.

     Actions are formalized as individual operators characterized by explicit preconditions, add effects, and delete effects that directly define how states transform.

     Sign in to read the full explanation

     Login / Signup

120. Question 116

     UGC NET June 2019

     Match List-I with List-II:

     List-I	List-II
     (a) Greedy best-first	(i) Minimal cost (p) + h(p)
     (b) Lowest cost-first	(ii) Minimal h(p)
     (c) A* algorithm	(iii) Minimal cost (p)

     1. A(a)-(i); (b)-(ii); (c)-(iii)
     2. B(a)-(iii); (b)-(ii); (c)-(i)
     3. C(a)-(i); (b)-(iii); (c)-(ii)
     4. D(a)-(ii); (b)-(iii); (c)-(i)

     Answer: (D) (a)-(ii); (b)-(iii); (c)-(i)

     Explanation

     Evaluation functions in informed search algorithms determine which path p is expanded first:

     (a)-(ii), (b)-(iii), (c)-(i)

     • (a) Greedy best-first — expands nodes solely based on the estimated distance to goal, minimizing h(p).
     • (b) Lowest cost-first — expands paths solely based on the accumulated path cost so far, minimizing cost(p).
     • (c) A algorithm* — balances actual path cost and estimated heuristic cost, minimizing cost(p) + h(p).

     Sign in to read the full explanation

     Login / Signup

121. Question 117

     UGC NET June 2019

     Consider the following:
     (a) Evolution
     (b) Selection
     (c) Reproduction
     (d) Mutation

     Which of the following are found in genetic algorithms?

     1. A(b), (c) and (d) only
     2. B(b) and (d) only
     3. C(a), (b), (c) and (d)
     4. D(a), (b) and (d) only

     Answer: (C) (a), (b), (c) and (d)

     Explanation

     Genetic algorithms model Darwinian natural evolution to solve complex optimization problems:

     • (a) Evolution — represents the generational adaptation of candidate solution populations.
     • (b) Selection — identifies the fittest chromosomes to survive and mate.
     • (c) Reproduction (Crossover) — recombines genetic material between pairs of parents to create offspring.
     • (d) Mutation — introduces stochastic perturbations in alleles to preserve population diversity.

     All four components are integral to genetic algorithms.

     Sign in to read the full explanation

     Login / Signup

122. Question 118

     UGC NET June 2019

     Consider the game tree given below:
     Here ○ and □ represents MIN and MAX nodes respectively.
     The value of the root node of the game tree is

     

     Game tree with alternating MIN and MAX nodes

     1. A4
     2. B7
     3. C11
     4. D12

     Answer: (B) 7

     Explanation

     Evaluating the minimax tree from bottom to top:

     1. At the lowest MIN level, each circular node selects the minimum of its square leaf children:
        First branch: min(7, 9) = 7
        Second branch: value 6
        Third branch: min(11, 12) = 11

     2. At the next level up (MAX), square nodes take the maximum of their child MIN nodes:
        Left subtree: max(7, 6) = 7
        Middle subtree: max(11, ...) = 11

     3. At the MIN level below the root, the left branch takes min(7, 11) = 7.

     4. At the root (MAX), comparing the subtrees evaluates to 7.

     Sign in to read the full explanation

     Login / Signup

123. Question 119

     UGC NET June 2019

     A fuzzy conjunction operator denoted as t(x, y) and a fuzzy disjunction operator denoted as s(x, y) form a dual pair if they satisfy the condition:

     1. At(x, y) = 1 - s(x, y)
     2. Bt(x, y) = s(1-x, 1-y)
     3. Ct(x, y) = 1 - s(1-x, 1-y)
     4. Dt(x, y) = s(1+x, 1+y)

     Answer: (C) t(x, y) = 1 - s(1-x, 1-y)

     Explanation

     In fuzzy set theory, t-norms (fuzzy conjunction) and s-norms (fuzzy disjunction) form dual pairs under generalized De Morgan's laws with respect to the standard fuzzy complement c(a) = 1 - a.

     The duality relation is defined as:
     c(t(x, y)) = s(c(x), c(y))
     1 - t(x, y) = s(1 - x, 1 - y)
     t(x, y) = 1 - s(1 - x, 1 - y).

     Sign in to read the full explanation

     Login / Signup

124. Question 120

     UGC NET June 2019

     K-mean clustering algorithm has clustered the given 8 observations into 3 clusters after 1st iteration as follows:
     C1: {(3, 3), (5, 5), (7, 7)}
     C2: {(0, 6), (6, 0), (3, 0)}
     C3: {(8, 8), (4, 4)}

     What will be the Manhattan distance for observation (4, 4) from cluster centroid C1 in second iteration?

     1. A2
     2. B√2
     3. C0
     4. D18

     Answer: (A) 2

     Explanation

     Calculating the centroid of cluster C1 and Manhattan distance:

     1. Cluster C1 contains {(3, 3), (5, 5), (7, 7)}.
        The centroid coordinates are the mean of these points:
        x_c1 = (3 + 5 + 7) / 3 = 15 / 3 = 5
        y_c1 = (3 + 5 + 7) / 3 = 15 / 3 = 5
        Centroid C1 = (5, 5).

     2. The Manhattan (L1) distance from point (4, 4) to centroid (5, 5) is:
        |x1 - x2| + |y1 - y2| = |4 - 5| + |4 - 5| = 1 + 1 = 2.

     Sign in to read the full explanation

     Login / Signup

125. Question 121

     UGC NET June 2019

     Reinforcement learning can be formalized in terms of ______ in which the agent initially only knows the set of possible ______ and the set of possible actions.

     1. AMarkov decision processes, objects
     2. BHidden states, objects
     3. CMarkov decision processes, states
     4. Dobjects, states

     Answer: (C) Markov decision processes, states

     Explanation

     Reinforcement learning models agent-environment interactions through the formal framework of Markov Decision Processes (MDPs).

     An MDP is formally defined by states, actions, transition dynamics, and rewards. In an unknown environment, the learning agent starts knowing only the set of possible states and the set of executable actions, discovering transition models and reward values through trial-and-error exploration.

     Sign in to read the full explanation

     Login / Signup

126. Question 122

     UGC NET June 2019

     Consider the following methods:
     M₁: Mean of maximum
     M₂: Centre of area
     M₃: Height method

     Which of the following is/are defuzzification method(s)?

     1. AOnly M₁
     2. BOnly M₁ and M₂
     3. COnly M₂ and M₃
     4. DM₁, M₂ and M₃

     Answer: (D) M₁, M₂ and M₃

     Explanation

     Defuzzification converts fuzzy membership sets produced by inference engines into crisp numerical values:

     • M1 (Mean of Maximum, MOM) calculates the average of all domain values that reach peak membership grade.
     • M2 (Centre of Area, COA / Centroid) calculates the geometric center of gravity across the aggregate membership curve.
     • M3 (Height method) evaluates crisp outputs by weighting the peak heights and center locations of individual fuzzy subsets.

     All three are recognized defuzzification techniques.

     Sign in to read the full explanation

     Login / Signup

127. Question 123

     UGC NET June 2019

     Let Aα₀ denotes the α-cut of a fuzzy set A at α₀. If α₁ < α₂, then

     1. AAα₁ ⊇ Aα₂
     2. BAα₁ ⊃ Aα₂
     3. CAα₁ ⊆ Aα₂
     4. DAα₁ ⊂ Aα₂

     Answer: (A) Aα₁ ⊇ Aα₂

     Explanation

     In fuzzy set theory, the alpha-cut A_alpha consists of all elements whose membership degree is at least alpha:
     A_alpha = { x ∈ X | μ_A(x) >= alpha }.

     When alpha1 < alpha2:
     Any element whose membership grade satisfies μ_A(x) >= alpha2 automatically satisfies μ_A(x) >= alpha1.
     Every element in A_alpha2 is therefore contained in A_alpha1.

     This establishes the superset inclusion relation A_alpha1 ⊇ A_alpha2.

     Sign in to read the full explanation

     Login / Signup

128. Question 124

     UGC NET June 2019

     Which of the following is an example of unsupervised neural network?

     1. ABack-propagation network
     2. BHebb network
     3. CAssociative memory network
     4. DSelf-organizing feature map

     Answer: (D) Self-organizing feature map

     Explanation

     Neural network architectures are classified by learning paradigm:

     • A. Back-propagation network — relies on supervised learning using error backpropagation computed from target output labels.
     • B. Hebb network — adjusts weights based on co-activation correlations.
     • C. Associative memory — associates patterns using stored vector matrices.
     • D. Self-organizing feature map (Kohonen SOM) — uses competitive unsupervised learning to map multi-dimensional input data onto a low-dimensional grid without supervisory feedback.

     Sign in to read the full explanation

     Login / Signup

129. Question 125

     UGC NET December 2018

     Which of the following is true for semi-dynamic environment ?

     1. AEnvironment and performance score, both change simultaneously.
     2. BThe environment may change while the agent is deliberating.
     3. CThe environment itself does not change with the passage of time but the agent’s performance score does.
     4. DEven if the environment changes with the passage of time while deliberating, the performance score does not change.

     Answer: (C) The environment itself does not change with the passage of time but the agent’s performance score does.

     Explanation

     A semi-dynamic environment is formally defined as one where the environment itself does not change while the agent is deliberating, but the agent's performance score decreases or changes with the passage of time.

     • In a fully dynamic environment, the physical world state changes while the agent ponders its move.
     • In a static environment, neither the environment nor the performance score changes during deliberation.
     • In a semi-dynamic environment (such as timed chess), the board state remains frozen until a move is made, but the clock runs down, altering the performance evaluation.
     • Therefore, Option C accurately captures the definition of a semi-dynamic environment.

     Sign in to read the full explanation

     Login / Signup
130. Practise Artificial Intelligence in a timed set

131. Question 126

     UGC NET December 2018

     Consider the following statements :

     S1 : A heuristic is admissible if it never overestimates the cost to reach the goal.
     S2 : A heuristic is monotonous if it follows triangle inequality property.

     Which one of the following is true referencing the above statements ?
     Choose the correct answer from the code given below :
     Code :

     1. ANeither of the statements S1 and S2 are true.
     2. BStatement S1 is false but statement S2 is true.
     3. CStatement S1 is true but statement S2 is false.
     4. DBoth the statements S1 and S2 are true.

     Answer: (D) Both the statements S1 and S2 are true.

     Explanation

     Both statements define core properties of heuristic functions used in informed search algorithms like A*.

     • Statement S1 is true because an admissible heuristic satisfies h(n) <= h*(n) for all nodes n, guaranteeing optimality.
     • Statement S2 is true because monotonicity (consistency) requires h(n) <= c(n, a, n') + h(n'), which is the triangle inequality in state space.
     • Because both definitions are standard in artificial intelligence, Option D is the correct choice.

     Sign in to read the full explanation

     Login / Signup

132. Question 127

     UGC NET December 2018

     An agent can improve its performance by

     1. AResponding
     2. BPerceiving
     3. CObserving
     4. DLearning

     Answer: (D) Learning

     Explanation

     In artificial intelligence, an agent improves its future performance specifically through its learning component.

     • In a learning agent architecture, the learning element takes feedback from the critic and modifies the performance element to produce better decisions over time.
     • Responding (Option A) is executing actions based on current rules.
     • Perceiving (Option B) and observing (Option C) provide raw sensory inputs from the environment but do not by themselves modify or improve agent performance.

     Sign in to read the full explanation

     Login / Signup

133. Question 128

     UGC NET December 2018

     Match List-I with List-II and choose the correct answer from the code:

     List-I	List-II
     (a) Greedy Best-First Search	(i) Selects a node for expansion if optimal path to that node has been found.
     (b) A* Search	(ii) Avoids substantial overhead associated with keeping the sorted queue of nodes.
     (c) Recursive Best-First Search	(iii) Suffers from excessive node generation.
     (d) Iterative-deepening A* Search	(iv) Time complexity depends on the quality of heuristic.

     Informed search algorithms and their characteristics

     1. A(iv) (i) (ii) (iii)
     2. B(i) (iv) (iii) (ii)
     3. C(iv) (iii) (ii) (i)
     4. D(i) (ii) (iii) (iv)

     Answer: (B) (i) (iv) (iii) (ii)

     Explanation

     Matching informed search algorithms with their core operational characteristics establishes the correct pairings.

     • (a) Greedy Best-First Search selects nodes for expansion solely based on heuristic evaluation h(n) estimating proximity to the goal, matching (i).
     • (b) A* Search incorporates both path cost g(n) and heuristic h(n), and its time complexity depends directly on the accuracy and quality of the heuristic function, matching (iv).
     • (c) Recursive Best-First Search (RBFS) operates in linear memory but repeatedly forgets and regenerates subtrees when backing up, suffering from excessive node generation, matching (iii).
     • (d) Iterative-deepening A* Search (IDA*) conducts successive depth-first searches bounded by f-cost cutoffs, avoiding the substantial overhead of maintaining an open sorted priority queue, matching (ii).

     Mapping:
     (a) -> (i), (b) -> (iv), (c) -> (iii), (d) -> (ii).

     • This code matches Option B: (i) (iv) (iii) (ii).

     Sign in to read the full explanation

     Login / Signup

134. Question 129

     UGC NET December 2018

     Consider the following statements related to AND-OR Search algorithm.

     S1 : A solution is a subtree that has a goal node at every leaf.
     S2 : OR nodes are analogous to the branching in a deterministic environment.
     S3 : AND nodes are analogous to the branching in a non-deterministic environment.

     Which one of the following is true referencing the above statements ?
     Choose the correct answer from the code given below :
     Code :

     1. AS1 – True, S2 – True, S3 – True
     2. BS1 – False, S2 – True, S3 – True
     3. CS1 – False, S2 – True, S3 – False
     4. DS1 – True, S2 – True, S3 – False

     Answer: (A) S1 – True, S2 – True, S3 – True

     Explanation

     All three statements are formal definitions of AND-OR search trees used for problem solving in nondeterministic environments.

     • Statement S1 is true because an AND-OR search solution is a conditional subtree where every leaf corresponds to a goal state.
     • Statement S2 is true because OR nodes represent the agent's alternative action choices, which corresponds to ordinary branching in a deterministic environment.
     • Statement S3 is true because AND nodes represent the environment's nondeterministic outcomes for a selected action, requiring the agent to have a contingency plan for every branch.
     • Therefore, all three statements S1, S2, and S3 are true (Option A).

     Sign in to read the full explanation

     Login / Signup

135. Question 130

     UGC NET December 2018

     Consider a vocabulary with only four propositions A, B, C and D. How many models are there for the following sentence ?

     ¬A ∨ ¬B ∨ ¬C ∨ ¬D

     1. A16
     2. B15
     3. C8
     4. D7

     Answer: (B) 15

     Explanation

     The sentence is false in exactly 1 model out of 2^4 = 16 total possible models, making it true in 15 models.

     • With 4 Boolean propositional symbols (A, B, C, D), there are 2^4 = 16 distinct truth assignments (models).
     • By De Morgan's Law, ¬A ∨ ¬B ∨ ¬C ∨ ¬D is logically equivalent to ¬(A ∧ B ∧ C ∧ D).
     • The conjunction (A ∧ B ∧ C ∧ D) is true in exactly one model, where all four variables are assigned True.
     • In that single model, the disjunction evaluates to False.
     • In all remaining 16 - 1 = 15 models, at least one variable is False, satisfying the disjunction.
     • Therefore, the sentence has 15 models (Option B).

     Sign in to read the full explanation

     Login / Signup

136. Question 131

     UGC NET December 2018

     Consider the following terminology and match List-I with List-II and choose the correct answer from the code:

     b = branching factor
     d = depth of the shallowest solution
     m = Maximum depth of the search tree
     l = depth limit

     List-I (Algorithms)	List-II (Space Complexity)
     (a) BFS search	(i) O(bd)
     (b) DFS search	(ii) O(b^d)
     (c) Depth-limited search	(iii) O(bm)
     (d) Iterative deepening search	(iv) O(bl)

     Search algorithms and their space complexities

     1. A(ii) (iii) (iv) (i)
     2. B(i) (iii) (iv) (ii)
     3. C(iii) (ii) (iv) (i)
     4. D(i) (iii) (iv) (ii)

     Answer: (A) (ii) (iii) (iv) (i)

     Explanation

     Matching search algorithms with their theoretical worst-case space complexities gives the standard artificial intelligence bounds.

     • (a) Breadth-First Search (BFS) stores every node in the frontier at depth d, requiring exponential space O(b^d), matching (ii).
     • (b) Depth-First Search (DFS) stores only the single active branch down to maximum depth m, requiring linear space O(bm), matching (iii).
     • (c) Depth-limited search is bounded by depth cutoff l, consuming linear space O(bl), matching (iv).
     • (d) Iterative deepening search (IDS) retains only a single branch down to current depth d, consuming linear space O(bd), matching (i).

     Mapping:
     (a) -> (ii), (b) -> (iii), (c) -> (iv), (d) -> (i).

     • This code corresponds to Option A: (ii) (iii) (iv) (i).

     Sign in to read the full explanation

     Login / Signup

137. Question 132

     UGC NET December 2018

     Consider the minimax game tree search shown in the diagram.

     What will be the value propagated at the root ?

     

     Minimax game tree search with alternating levels

     1. A4
     2. B5
     3. C6
     4. D3

     Answer: (C) 6

     Explanation

     Evaluating the game tree bottom-up with alternating MAX and MIN choices yields a root value of 6.

     1. Level 3 (lowest MAX nodes above leaves):

     • Subtree 1 leaves: 4, 6, 2, 6 -> max(4, 6, 2, 6) = 6
     • Subtree 2 leaves: 3, 9, 5, 2 -> max(3, 9, 5, 2) = 9
     • Subtree 3 leaves: 7, 3, 1, 7 -> max(7, 3, 1, 7) = 7
     • Subtree 4 leaves: 2, 4, 6, 3 -> max(2, 4, 6, 3) = 6

     1. Level 2 (intermediate MIN nodes):

     • Left MIN node chooses between Subtrees 1 and 2: min(6, 9) = 6
     • Right MIN node chooses between Subtrees 3 and 4: min(7, 6) = 6

     1. Level 1 (root MAX node):

     • Root MAX chooses between left and right MIN nodes: max(6, 6) = 6.

     • Therefore, the value propagated at the root is 6 (Option C).

     Sign in to read the full explanation

     Login / Signup

138. Question 133

     UGC NET December 2018

     A full joint distribution for the Toothache, Cavity and Catch is given in the table:

     	Toothache Catch	Toothache ¬Catch	¬Toothache Catch	¬Toothache ¬Catch
     Cavity	0·108	0·012	0·072	0·008
     ¬Cavity	0·016	0·064	0·144	0·576

     What is the probability of Cavity, given evidence of Toothache?

     1. A<0·2, 0·8>
     2. B<0·6, 0·8>
     3. C<0·6, 0·4>
     4. D<0·4, 0·8>

     Answer: (C) <0·6, 0·4>

     Explanation

     Conditioning on the evidence of Toothache and normalizing across Cavity states produces distribution <0.6, 0.4>.

     From the joint probability table, summing under the condition Toothache = true:

     1. For Cavity = true:
        P(Cavity ∧ Toothache) = P(Cavity, Toothache, Catch) + P(Cavity, Toothache, ¬Catch)
        = 0.108 + 0.012 = 0.120.

     2. For Cavity = false (¬Cavity):
        P(¬Cavity ∧ Toothache) = P(¬Cavity, Toothache, Catch) + P(¬Cavity, Toothache, ¬Catch)
        = 0.016 + 0.064 = 0.080.

     3. Total marginal probability P(Toothache):
        P(Toothache) = 0.120 + 0.080 = 0.200.

     4. Normalizing the conditional probabilities:
        P(Cavity | Toothache) = 0.120 / 0.200 = 0.6.
        P(¬Cavity | Toothache) = 0.080 / 0.200 = 0.4.

     • The conditional probability distribution vector is <0.6, 0.4> (Option C).

     Sign in to read the full explanation

     Login / Signup

Frequently asked questions

More Computer Science PYQs by topic

More Computer Science practice