Question 1 of 68Introduction to AI and intelligent agent
The term 'Artificial Intelligence' was formally coined at which event?
AThe Turing Test proposal
BThe 1956 Dartmouth Conference
CThe first computer science degree program
DThe invention of the perceptron
Answer is hidden
Question 2 of 68Introduction to AI and intelligent agent
An agent's actuators are used to:
APerceive the environment
BAct upon the environment
CStore past percepts only
DCompute a utility function only
Answer is hidden
Question 3 of 68Introduction to AI and intelligent agent
A simple reflex agent selects its actions based on:
AA model of the world and future goals
BOnly the current percept, using condition-action rules
CA utility function
DCommunication with other agents
Answer is hidden
Question 4 of 68Introduction to AI and intelligent agent
Which type of agent maintains an internal state to handle partial observability?
ASimple reflex agent
BModel-based reflex agent
CRandom agent
DNone of the agent types
Answer is hidden
Question 5 of 68Introduction to AI and intelligent agent
A utility-based agent is most useful when:
AThere is only one possible goal
BMultiple goals conflict and a trade-off/preference must be chosen
CThe environment is always deterministic
DNo actions are available
Answer is hidden
Question 6 of 68Introduction to AI and intelligent agent
An environment where the outcome of an action is not fully predictable is called:
ADeterministic
BStochastic
CStatic
DFully observable
Answer is hidden
Question 7 of 68Introduction to AI and intelligent agent
An environment that can change while the agent is still deciding on an action is called:
AStatic
BDynamic
CDeterministic
DSingle-agent
Answer is hidden
Question 8 of 68Introduction to AI and intelligent agent
If an agent's sensors provide complete information about the current state, the environment is:
APartially observable
BFully observable
CStochastic
DMulti-agent
Answer is hidden
Question 9 of 68Introduction to AI and intelligent agent
The 'acting humanly' perspective of AI is closely associated with:
AThe Turing Test
BBayesian networks
CGenetic algorithms
DBackpropagation
Answer is hidden
Question 10 of 68Introduction to AI and intelligent agent
A goal-based agent differs from a simple reflex agent mainly by:
AHaving no actuators
BConsidering future consequences of actions to achieve a goal
CIgnoring all percepts
DNever using condition-action rules
Answer is hidden
Question 11 of 68Problem solving and searching techniques
A Constraint Satisfaction Problem is defined by:
AStates and actions only
BVariables, domains, and constraints
CA utility function only
DA heuristic function only
Answer is hidden
Question 12 of 68Problem solving and searching techniques
Which uninformed search strategy explores the search tree level by level?
ADFS
BBFS
CHill climbing
DA* search
Answer is hidden
Question 13 of 68Problem solving and searching techniques
Depth-First Search generally uses:
AMore memory than BFS
BLess memory than BFS
CExactly the same memory as BFS
DNo memory at all
Answer is hidden
Question 14 of 68Problem solving and searching techniques
Iterative Deepening Search combines the advantages of:
AOnly DFS
BOnly BFS
CBFS's completeness/optimality with DFS's low memory usage
DMinimax and alpha-beta pruning
Answer is hidden
Question 15 of 68Problem solving and searching techniques
A* search selects the next node to expand based on:
AOnly the heuristic h(n)
BOnly the path cost g(n)
Cf(n) = g(n) + h(n)
DA random selection
Answer is hidden
Question 16 of 68Problem solving and searching techniques
A* search is guaranteed to find the optimal solution when the heuristic is:
AAlways zero
BAdmissible (never overestimates the true cost)
CAlways an overestimate
DIgnored entirely
Answer is hidden
Question 17 of 68Problem solving and searching techniques
Hill climbing search can fail to find the global optimum because it may get stuck at a:
AGlobal maximum
BLocal maximum
CStarting state only
DGoal state
Answer is hidden
Question 18 of 68Problem solving and searching techniques
Simulated annealing differs from basic hill climbing by:
ANever accepting a worse move
BOccasionally accepting a worse move to escape local optima
CIgnoring the current state entirely
DOnly working on adversarial games
Answer is hidden
Question 19 of 68Problem solving and searching techniques
The Minimax algorithm is used primarily for:
ASingle-agent path planning
BTwo-player adversarial (zero-sum) games
CClustering unlabeled data
DTraining neural networks
Answer is hidden
Question 20 of 68Problem solving and searching techniques
Alpha-beta pruning improves minimax by:
AChanging the final decision to a suboptimal one
BSkipping branches that cannot affect the final decision, without changing the result
CRemoving the need for a game tree
DMaking the search less accurate
Answer is hidden
Question 21 of 68Knowledge representation
A semantic network represents knowledge using:
AOnly numeric probabilities
BNodes (concepts) connected by labeled edges (relationships)
COnly if-then rules
DGenetic algorithm chromosomes
Answer is hidden
Question 22 of 68Knowledge representation
A frame represents knowledge using:
ASlots and fillers
BOnly logical connectives
CA game tree
DA neural network layer
Answer is hidden
Question 23 of 68Knowledge representation
A tautology in propositional logic is a formula that is:
AAlways false
BAlways true, regardless of interpretation
CSometimes true, sometimes false
DUndefined
Answer is hidden
Question 24 of 68Knowledge representation
Resolution as an inference technique works by:
ADirectly guessing the answer
BConverting statements to clausal form and deriving a contradiction
CIgnoring all premises
DOnly working with fuzzy logic
Answer is hidden
Question 25 of 68Knowledge representation
The symbol โ in First-Order Predicate Logic represents:
AExistential quantification ('there exists')
BUniversal quantification ('for all')
CLogical negation
DLogical implication
Answer is hidden
Question 26 of 68Knowledge representation
Unification in FOPL is used to:
AFind a substitution that makes two expressions identical
BCompute a truth table
CBuild a semantic network
DTrain a neural network
Answer is hidden
Question 27 of 68Knowledge representation
Bayes' Rule is expressed as:
AP(A|B) = P(A) + P(B)
BP(A|B) = P(B|A)P(A)/P(B)
CP(A|B) = P(A)/P(B) only
DP(A|B) = P(B) - P(A)
Answer is hidden
Question 28 of 68Knowledge representation
A Bayesian network is structured as a:
ADirected acyclic graph (DAG) of variables and dependencies
BBinary search tree
CSimple linked list
DFully connected undirected graph only
Answer is hidden
Question 29 of 68Knowledge representation
Reasoning in belief (Bayesian) networks typically involves computing:
AThe shortest path between nodes
BThe posterior probability of variables given evidence
CA minimax value
DA fitness function
Answer is hidden
Question 30 of 68Knowledge representation
Which of these is an issue commonly identified in knowledge representation?
AChoosing the correct programming language
BChoosing the right granularity and structure for representing knowledge
CDetermining CPU clock speed
DSelecting a network topology
Answer is hidden
Question 31 of 68Expert system and natural language processing
The component of an expert system that applies rules to derive new conclusions is the:
AKnowledge base
BInference engine
CUser interface
DExplanation facility
Answer is hidden
Question 32 of 68Expert system and natural language processing
'Knowing how to ride a bicycle' is an example of:
ADeclarative knowledge
BProcedural knowledge
CSemantic knowledge only
DNo knowledge type
Answer is hidden
Question 33 of 68Expert system and natural language processing
Which expert system component explains the reasoning behind its conclusions?
AKnowledge base
BExplanation facility
CKnowledge acquisition module
DInference engine only
Answer is hidden
Question 34 of 68Expert system and natural language processing
The step in NLP that involves breaking text into individual words/tokens is called:
ASemantic analysis
BLexical analysis
CPragmatic analysis
DDiscourse integration
Answer is hidden
Question 35 of 68Expert system and natural language processing
Generating human-readable text from structured data is the task of:
ANatural Language Understanding (NLU)
BNatural Language Generation (NLG)
CMachine vision
DGenetic algorithms
Answer is hidden
Question 36 of 68Expert system and natural language processing
Ambiguity in language (a word or sentence having multiple meanings) is considered a:
ABenefit of NLP systems
BMajor challenge in NLP
CNon-issue in NLP
DOnly a speech recognition issue
Answer is hidden
Question 37 of 68Expert system and natural language processing
Which stage of machine vision involves isolating regions of interest in an image?
AImage acquisition
BSegmentation
CPreprocessing only
DFeature extraction only
Answer is hidden
Question 38 of 68Expert system and natural language processing
Robotics combines AI with which other disciplines?
AOnly pure mathematics
BMechanical engineering and electronics
COnly linguistics
DOnly databases
Answer is hidden
Question 39 of 68Expert system and natural language processing
Knowledge acquisition in expert system development refers to:
ADeleting outdated rules
BExtracting and structuring expert knowledge into the knowledge base
CRunning the inference engine
DOnly user interface design
Answer is hidden
Question 40 of 68Expert system and natural language processing
A knowledge engineer's primary role in expert system development is to:
AWrite the final report only
BWork with domain experts to elicit and encode their knowledge
COnly design the graphical interface
DOptimize hardware performance
Answer is hidden
Question 41 of 68Machine learning
Supervised learning requires training data that is:
AUnlabeled
BLabeled with known input-output pairs
CRandomly generated only
DAlways numeric
Answer is hidden
Question 42 of 68Machine learning
Which learning paradigm finds patterns in unlabeled data (e.g., clustering)?
ASupervised learning
BUnsupervised learning
CReinforcement learning
DGenetic learning
Answer is hidden
Question 43 of 68Machine learning
In reinforcement learning, an agent learns by:
AUsing only labeled datasets
BReceiving rewards or penalties from interacting with an environment
CClustering data points
DApplying fuzzy logic rules
Answer is hidden
Question 44 of 68Machine learning
A decision tree typically chooses attribute splits based on:
ARandom selection
BInformation gain (or entropy reduction)
CFuzzy membership values
DGenetic crossover
Answer is hidden
Question 45 of 68Machine learning
The Naive Bayes classifier assumes that features are:
APerfectly correlated
BConditionally independent given the class
CAlways continuous
DIrrelevant to classification
Answer is hidden
Question 46 of 68Machine learning
In a Fuzzy Inference System, the process of converting a crisp input into a fuzzy value is called:
ADefuzzification
BFuzzification
CSelection
DCrossover
Answer is hidden
Question 47 of 68Machine learning
Converting the fuzzy output of an inference system back into a crisp value is called:
AFuzzification
BDefuzzification
CMutation
DEncoding
Answer is hidden
Question 48 of 68Machine learning
Which of these is a well-known fuzzy inference method?
AMamdani method
BBackpropagation method
CMinimax method
DA* method
Answer is hidden
Question 49 of 68Machine learning
In a Genetic Algorithm, the operator that combines parts of two parent chromosomes is called:
ASelection
BCrossover
CMutation
DFitness evaluation
Answer is hidden
Question 50 of 68Machine learning
The fitness function in a Genetic Algorithm is used to:
ASelect the encoding scheme
BEvaluate how good a candidate solution is
CPerform defuzzification
DConvert labels to features
Answer is hidden
Question 51 of 68Neural networks
The McCulloch-Pitts neuron is best described as:
AA modern deep learning architecture
BAn early binary threshold model of a neuron
CA fuzzy inference system
DA genetic algorithm operator
Answer is hidden
Question 52 of 68Neural networks
In an artificial neuron, the activation function is applied to:
AThe raw input values only
BThe weighted sum of the inputs
CThe learning rate
DThe fitness function
Answer is hidden
Question 53 of 68Neural networks
Which activation function outputs values strictly between 0 and 1?
AReLU
BSigmoid
CStep function only
DLinear function
Answer is hidden
Question 54 of 68Neural networks
A single-layer Perceptron can correctly classify data that is:
ANon-linearly separable only
BLinearly separable
CAlways randomly distributed
DNever classifiable
Answer is hidden
Question 55 of 68Neural networks
Gradient descent updates a neural network's weights by moving:
AIn a random direction
BIn the direction that increases the error
CIn the direction that most reduces the error (negative gradient)
DOnly after training is complete
Answer is hidden
Question 56 of 68Neural networks
Hebbian learning is summarized by the principle:
A'The strongest neuron wins'
B'Neurons that fire together, wire together'
C'Error must always be zero'
D'Only output neurons learn'
Answer is hidden
Question 57 of 68Neural networks
The Adaline model differs from the basic Perceptron mainly by using:
AA step activation function during training
BA linear activation function trained with the delta rule
CNo activation function at all
DOnly unsupervised learning
Answer is hidden
Question 58 of 68Neural networks
A Multilayer Perceptron (MLP) is capable of solving problems that a single-layer perceptron cannot, mainly because it has:
AMore output neurons only
BOne or more hidden layers enabling non-linear decision boundaries
CNo activation function
DOnly linear connections
Answer is hidden
Question 59 of 68Neural networks
The Backpropagation algorithm trains a neural network by:
APropagating error backward through the network to update weights via gradient descent
BRandomly guessing weights until correct
COnly adjusting the input layer
DIgnoring the output error
Answer is hidden
Question 60 of 68Neural networks
A Hopfield network is primarily used as a(n):
AFeedforward classifier only
BAssociative (content-addressable) memory
CGenetic algorithm encoder
DFuzzy inference engine
Answer is hidden
Question 61 of 68Information Theory & Signals
The Fourier Transform is used to:
AConvert a signal from the time domain to the frequency domain
BConvert a signal from the frequency domain to the time domain
CCompress digital signals
DEncrypt data transmissions
Answer is hidden
Question 62 of 68Information Theory & Signals
When a square wave is applied to an integrator op-amp circuit, the output is:
ASquare wave
BSine wave
CTriangle wave
DSawtooth wave
Answer is hidden
Question 63 of 68Information Theory & Signals
The Z-transform is primarily used for:
AAnalysing continuous-time signals
BAnalysing discrete-time signals and systems
CImage compression
DFrequency modulation
Answer is hidden
Question 64 of 68Information Theory & Signals
Interleaving in digital communications provides what type of diversity?
ATime diversity
BFrequency diversity
CSpace diversity
DPolarization diversity
Answer is hidden
Question 65 of 68Information Theory & Signals
In FDM, increasing the modulation index results in:
ADecreased bandwidth requirements
BIncreased bandwidth requirements
CNo change in bandwidth
DIncreased signal-to-noise ratio without bandwidth change
Answer is hidden
Question 66 of 68Information Theory & Signals
Adding zeros to the end of a discrete signal before taking its DFT is called:
AZero Padding
BWindowing
CDecimation
DInterpolation
Answer is hidden
Question 67 of 68Information Theory & Signals
In TDM (Time Division Multiplexing), the component essential for recovering individual channels at the receiver is:
AEqualizer
BFrame synchronization
CBandwidth filter
DDemodulator
Answer is hidden
Question 68 of 68Information Theory & Signals
In a common-collector (emitter follower) BJT amplifier, the phase relationship between input and output is:
A180 degrees (inverted)
B90 degrees
C0 degrees (in phase)
D45 degrees
Answer is hidden