wayground logo

Free Printable Worksheets

Font size

S
M
L
XL
Worksheets

Introduction to Artificial Neural Networks

Total questions: 124

Worksheet time: 1hrs 2mins

Name
Class
Date
1.

Which statement best defines an Artificial Neural Network in modern computing?

a)

A computational model inspired by biological neurons

b)

A hardware-only solution for parallel processors

c)

A rule-based expert system using logic trees

d)

A statistical test for comparing two datasets

2.

Why do Artificial Neural Networks occupy a central role in machine learning?

a)

They generalize across vast application areas

b)

They eliminate the need for training data

c)

They only work for small, niche tasks

d)

They replace all other AI techniques entirely

3.

In the context of ANNs, what does Explicit Learning most likely refer to?

a)

Clear, structured learning with defined targets

b)

Unsupervised discovery without labeled data

c)

Random exploration through trial-and-error

d)

Implicit pattern formation without guidance

4.

Which model by McCulloch and Pitts in 1943 provided a mathematical basis for early neural networks?

a)

Threshold Logic Model

b)

Linear Regression Model

c)

Bayesian Inference Model

d)

Hidden Markov Model

5.

Hebbian Learning is best characterized as which learning paradigm?

a)

Supervised learning rules

b)

Semi‑supervised learning rules

c)

Reinforcement learning rules

d)

Unsupervised learning rules

6.

Who is credited with inventing the perceptron in 1958?

a)

Paul Werbos

b)

Donald Hebb

c)

Warren McCulloch

d)

Frank Rosenblatt

7.

What core operation did early perceptrons learn across two layers?

a)

Addition and subtraction

b)

Convolution and pooling

c)

Multiplication and division

d)

Exponentiation and log

8.

Which researcher formulated back‑propagation as a learning algorithm in 1975?

a)

Seymour Papert

b)

Paul Werbos

c)

Alan Turing

d)

Marvin Minsky

9.

Research in neural networks declined between 1960 and 1975 largely due to which paired issues?

a)

Limited data and poor sensors

b)

Weak theories and ethical bans

c)

Costly hardware and few researchers

d)

Simple models and slow computers

10.

Which development in the 1980s–1990s most enabled the resurgence of ANN research?

a)

Quantum processors widespread

b)

Analog circuit miniaturization

c)

Parallel computing processors growth

d)

Optical networking standardization

11.

Farley and Clark’s work in 1954 used computation machines and simulations at which institution?

a)

Stanford University

b)

Carnegie Mellon University

c)

Harvard University

d)

Massachusetts Institute of Technology

12.

In the biological neuron diagram, which structure primarily receives signals from other neurons?

a)

Axon extending from the cell body

b)

Nucleus located within the soma

c)

Synapse terminals at the axon tip

d)

Dendrites branching near the soma

13.

Which pairing best maps a biological component to its artificial neuron counterpart?

a)

Dendrites to input features x

b)

Nucleus to adjustable weights w

c)

Axon to activation function g

d)

Synapse to network topology

14.

In the artificial neuron diagram, what is computed immediately before the activation function block?

a)

Bias subtraction from output

b)

Weighted sum Σ of inputs

c)

Gradient of loss function

d)

Binary thresholding decision

15.

Which statement correctly describes weights in an artificial neuron?

a)

Scalars that scale each input

b)

Outputs that store final predictions

c)

Functions that compute gradients

d)

Hyperparameters that define topology

16.

Which set lists valid input types for an artificial neuron?

a)

Continuous, discrete, categorical

b)

Nominal, ordinal, ratio

c)

Text, image, audio

d)

Integer, real, binary

17.

What role does the activation function play in an artificial neuron?

a)

Normalizes inputs before summation

b)

Routes signals between layers

c)

Transforms summed input to output

d)

Stores weights during training

18.

Which description best matches ANN topology?

a)

Range of valid input formats

b)

Magnitude of each weight parameter

c)

Type of activation function used

d)

Arrangement of neurons and connections

19.

Which activation function outputs exactly two levels, switching at x = 0 as shown in the diagram?

a)

Symmetric linear function

b)

Binary hard-limit function

c)

Bipolar hard-limit function

d)

Continuous sigmoid function

20.

In the plotted symmetric activation, f(x) equals x. What property does this function exhibit across positive and negative inputs?

a)

Saturating at large magnitudes

b)

Discrete step transitions

c)

Even symmetry around y-axis

d)

Odd symmetry around origin

21.

Which curve approaches 0 and 1 asymptotically and is given by f(x)=11+exf(x)=\frac{1}{1+e^{-x}} ?

a)

Binary step function

b)

Logistic sigmoid function

c)

Piecewise stairs function

d)

Hyperbolic tangent function

22.

Which activation maps negative inputs to −1 and positive inputs to +1 as depicted?

a)

Sigmoid logistic function

b)

Linear symmetric function

c)

Binary hard-limit function

d)

Bipolar hard-limit function

23.

The tanh activation shown has formula (exex)/(ex+ex)(e^x − e^{−x})/(e^x + e^{−x}) . What is its output range?

a)

From −1 to +1

b)

From 0 to +1

c)

From −∞ to +∞

d)

From 0 to +2

24.

Which activation illustrates multiple discrete plateaus labeled a, b, c with outputs 0,1,2,3?

a)

Symmetric linear function

b)

Continuous tanh function

c)

Multi-level steps function

d)

Binary threshold function

25.

Which ANN topology connects inputs directly to outputs without hidden layers, as drawn?

a)

Single-layer neural network

b)

Feedback single-layer network

c)

Multi-layer neural network

d)

Self-organizing network

26.

Which topology shows recurrent connections from outputs back into the network, enabling feedback?

a)

Purely linear perceptron

b)

Feedback single-layer network

c)

Self-organizing network

d)

Multi-layer feedforward network

27.

Which ANN property enables multiple processing units to work simultaneously to increase speed?

a)

Parallelism and synchronization

b)

Distributed memory storage

c)

Generalization capability

d)

Ease of construction

28.

What does adaptation in ANNs primarily allow the network to do?

a)

Store data permanently

b)

Ignore input variations

c)

Adjust to problem-specific data

d)

Follow fixed program steps

29.

Distributed memory in ANNs means information is stored how?

a)

Only during training sessions

b)

Externally in databases

c)

Across many processing elements

d)

In one central unit

30.

Generalization in ANNs refers to the ability to do what after training?

a)

Memorize training samples

b)

Perform on unseen but related data

c)

Require retraining for each case

d)

Eliminate all input noise

31.

Fault tolerance in ANNs helps primarily with which situation?

a)

Precise feature engineering

b)

Long-term storage capacity

c)

Errors in inputs or processing

d)

Sequential execution speed

32.

Problem definition in ANNs is important because the network learns from what?

a)

Purely manual tuning

b)

Prewritten program rules

c)

A set of modeled inputs

d)

Random unstructured inputs

33.

Handling noisy data in ANNs typically involves which capability?

a)

Reject all imperfect samples

b)

Disable learning on noise

c)

Infer patterns despite errors

d)

Require complete datasets

34.

Ease of construction and learning in ANNs is often due to what?

a)

Highly customized code

b)

Standard algorithms and rules

c)

Manual configuration only

d)

Hardware-only solutions

35.

Parallelism in ANNs contrasts most directly with which architecture style?

a)

Von Neumann sequential processing

b)

Event-driven reactive systems

c)

Quantum probabilistic computing

d)

Pipeline vector processing

36.

Which property helps ANNs avoid overreliance on exact training samples?

a)

Generalization capability

b)

Fault tolerance

c)

Distributed memory

d)

Adaptation ability

37.

If some feature values are missing, which ANN property still enables useful outputs?

a)

Fault tolerance

b)

Problem definition

c)

Parallelism

d)

Ease of construction

38.

Why is clear problem definition critical before training an ANN?

a)

It removes need for data collection

b)

It allows fixed rule execution

c)

It sets algorithmic programming steps

d)

It determines network architecture

39.

Adaptation in ANNs reduces dependence on what traditional approach?

a)

Rule-based programming

b)

Gradient descent updates

c)

Data augmentation schemes

d)

Batch normalization layers

40.

Distributed memory contributes to robustness because information is what?

a)

Redundant across units

b)

Compressed in one node

c)

Encrypted after training

d)

Discarded when noisy

41.

Generalization is commonly assessed by performance on which dataset?

a)

Training set only

b)

Validation or test set

c)

Randomized noise samples

d)

Memory recall set

42.

Fault tolerance means the network can do what when inputs contain errors?

a)

Require manual fixes

b)

Retrain from scratch

c)

Correct or compensate

d)

Stop processing entirely

43.

Which pairing best matches property to benefit: Parallelism → ?

a)

Better storage through redundancy

b)

Higher speed through concurrency

c)

Simpler coding via templates

d)

Improved accuracy via rules

44.

Which scenario best illustrates adaptation ability?

a)

Data is stored in one memory

b)

Inputs are strictly noise-free

c)

Weights update to fit task data

d)

A fixed program runs identical steps

45.

Which scenario best demonstrates handling noisy data?

a)

Model discards all corrupted rows

b)

Model stops on any outlier

c)

Model infers despite missing values

d)

Model requires perfect labels

46.

Ease of construction in ANNs often results from availability of what?

a)

Hand-written case rules

b)

Standardized learning procedures

c)

Manual feature extraction

d)

Custom hardware only

47.

Which statement best describes generalization in artificial neural networks?

a)

Memorizing all training examples exactly

b)

Optimizing weights without any data

c)

Stopping training when loss reaches zero

d)

Learning with minimal training data size

48.

According to the diagram, how does ANN performance typically change as the learning data size increases?

a)

Gradually increases over larger data

b)

Rapidly decreases with more data

c)

Remains constant regardless of data

d)

Oscillates unpredictably with size

49.

In supervised learning, what indicates that the learning process should stop?

a)

Bias value equals learning rate

b)

Clusters fully formed without labels

c)

Weights become stable on inputs

d)

Network learns input–output mappings

50.

Which learning type relies solely on input data without target outputs?

a)

Unsupervised learning approach

b)

Supervised learning approach

c)

Reinforcement learning approach

d)

Self-organization approach

51.

Self-organization in ANNs is most similar to which learning type?

a)

Unsupervised learning methods

b)

Semi-supervised methods

c)

Active learning methods

d)

Supervised learning methods

52.

Which technique is commonly associated with self-organization processes?

a)

Clustering of input patterns

b)

Gradient descent on loss

c)

Labeling with target outputs

d)

Backpropagation with labels

53.

Which learning rule updates weights proportionally to the product of input and output?

a)

Hebb rule formulation

b)

Widrow–Hoff formulation

c)

Competitive rule formulation

d)

Generalized delta formulation

54.

The Widrow–Hoff rule adjusts weights using which error term?

a)

Difference between target and output

b)

Difference between bias and weight

c)

Sum of input and output values

d)

Product of input and output values

55.

In the competitive rule, the update depends on which comparison?

a)

Input minus current weight

b)

Target minus current output

c)

Bias minus learning rate

d)

Momentum minus decay factor

56.

The Generalized Delta Rule uses which quantity to scale the input during weight updates?

a)

Fixed bias threshold

b)

Output normalization value

c)

Delta of the neuron j

d)

Random momentum term

57.

Adding momentum to the GDR introduces which extra term in the update?

a)

Divide by eta learning rate

b)

Add beta times bias

c)

Subtract alpha times weight

d)

Multiply gamma times input

58.

What does the learning rate (η) control during training?

a)

Number of hidden neurons

b)

Magnitude of bias values

c)

Choice of activation function

d)

Step size of weight changes

59.

In ANN models, what role does the bias or threshold (θ) primarily play?

a)

Replaces the learning rate

b)

Shifts neuron activation level

c)

Sets training data size

d)

Determines output labels

60.

Who is credited with developing the back-propagation algorithm and in which year was it introduced?

a)

Paul Werbos in 1975

b)

Geoff Hinton in 1986

c)

Yann LeCun in 1998

d)

Alan Turing in 1950

61.

Which learning paradigm best describes training in a Back-Propagation Neural Network?

a)

Unsupervised learning

b)

Reinforcement learning

c)

Supervised learning

d)

Self-supervised learning

62.

What is the primary role of non-linear activation functions in multi-layer networks?

a)

Reduce network connectivity

b)

Enable linear separability

c)

Allow complex function approximation

d)

Fix weight initialization issues

63.

In the forward-backward process of BP-NN, what happens during the backward phase?

a)

Inputs propagate to outputs

b)

Errors propagate to update weights

c)

Bias terms are removed entirely

d)

Weights are frozen for stability

64.

Which topology characteristic is depicted for BP-NN in the diagram?

a)

Single-layer perceptron

b)

Sparse connections only

c)

Recurrent feedback loops

d)

Full connection between layers

65.

What does the learning rate η control during training?

a)

Magnitude of weight updates

b)

Number of hidden layers

c)

Size of training dataset

d)

Choice of activation function

66.

Which statement best describes the momentum term α in gradient-based updates?

a)

Adds random noise to gradients

b)

Accumulates past updates to speed convergence

c)

Sets an absolute threshold for activation

d)

Fixes weights to prevent overfitting

67.

In a BP-NN with A inputs, B hidden nodes, and C outputs, what is the correct shape of the first-layer weight matrix W1?

a)

A×B matrix mapping inputs to hidden

b)

C×B matrix mapping outputs to hidden

c)

B×A matrix mapping hidden to inputs

d)

A×C matrix mapping inputs to outputs

68.

Which parameter controls the magnitude of weight updates during training in back-propagation?

a)

Activation function

b)

Batch size P

c)

Learning rate η

d)

Output count C

69.

Hidden node net input v is typically computed as which operation?

a)

Cumulative product of inputs

b)

Softmax over inputs X

c)

Element-wise product X⊙W1

d)

Weighted sum W1·X

70.

For a sigmoid unit, which formula gives the hidden node activation Hj from net input v?

a)

Hj = tanh(v)

b)

Hj = 11+ev\frac{1}{1+e^{−v}}

c)

Hj = ReLU(v)

d)

Hj=v2Hj = v^2

71.

Given hidden activations H and second-layer weights W2, how is the output node net input computed?

a)

Difference between first two hidden nodes

b)

Max pooling of hidden nodes

c)

Average of hidden activations

d)

Weighted sum over hidden nodes

72.

Which expression represents the error across C outputs for one sample when using root-sum-squared error?

a)

error = sqrt(Σk(YkOk)2)sqrt(Σ_k (Yk − Ok)^2)

b)

error = Σ_k (Yk − Ok)

c)

error = mean(Ok)

d)

error = ΣkOk2Σ_k Ok^2

73.

What is the output-layer delta term δ2_k for a sigmoid output unit?

a)

Ok(1−Ok)(Yk−Ok)

b)

Ok+Yk

c)

Ok(1+Ok)(Yk+Ok)

d)

Ok−Yk

74.

Which term propagates error to the hidden layer in back-propagation?

a)

P = number of samples

b)

η = Σ_i Xi

c)

Zk = Ok − Yk

d)

δ1_k = Zk(1−Zk) Σ_j δ2_j W2_{k,j}

75.

During weight update for W2, which formula is used for a single sample?

a)

W2_{i,j} ← W2_{i,j} − η Zk

b)

W2_{i,j} ← W2_{i,j} + η Ok

c)

W2_{i,j} ← W2_{i,j} − η Xi

d)

W2_{i,j} ← W2_{i,j} + η δ2_j H_i

76.

During weight update for W1, which quantity multiplies the input Xi?

a)

Error scalar only

b)

Output-layer delta δ2_k

c)

Output value Ok

d)

Hidden-layer delta δ1_j

77.

What is the correct iterative structure for training with back-propagation?

a)

Randomly initialize each epoch only

b)

Update weights before computing outputs

c)

While error not acceptable, loop over samples

d)

Compute outputs once, then end

78.

Which step occurs first in the BP-NN algorithm?

a)

Calculate hidden-layer deltas

b)

Compute output node values

c)

Initialize parameters and activation

d)

Save the last weight values

79.

What does random initialization of W1 and W2 primarily prevent?

a)

Symmetry among neurons

b)

Need for activation functions

c)

Exact gradient computation

d)

Overfitting immediately

80.

If the error is acceptable, what should the algorithm do next?

a)

End the while loop

b)

Increase the learning rate

c)

Reset the weights

d)

Recompute deltas repeatedly

81.

Which part of the algorithm saves the trained model state?

a)

Save the last weight values

b)

Calculate δ2 at output

c)

Initialize parameter A

d)

Compute hidden activations

82.

In this worked example, which activation function is used for hidden and output nodes when computing h1, h2, o1, and o2?

a)

Hyperbolic tangent function

b)

Softmax normalization

c)

Logistic sigmoid function

d)

Rectified linear unit

83.

Given input sample [1 0 0] and W1 weights shown, what is the computed value of h1 using the activation applied to sum1 = 0.2?

a)

h1 ≈ 0.4502 after activation

b)

h1 ≈ 0.7311 after activation

c)

h1 ≈ 0.5498 after activation

d)

h1 ≈ 0.5915 after activation

84.

For the same sample, Sum2 at the output layer is 0.805 from hidden activations and W2. After activation, what is o2 approximately?

a)

o2 ≈ 0.551 using activation

b)

o2 ≈ 0.37 using activation

c)

o2 ≈ 0.691 using activation

d)

o2 ≈ 0.525 using activation

85.

The desired output is [1 0], while the actual output is [0.5915 0.691]. Which statement best describes the error acceptance decision shown?

a)

The error is acceptable and training stops

b)

The error is not acceptance so continue training

c)

The error equals zero so weights remain

d)

The error is averaged then discarded

86.

Using δ values at the output layer, which expression matches δ2 for o1 given o1 = 0.5915?

a)

δ2 = o1(1−o1)(1−desired1)

b)

δ2 = o1(1−o1)(desired1−o1)

c)

δ2 = o1(1−o1)(1−o1)

d)

δ2 = o1(1−o1)(o1−desired1)

87.

Which formula correctly computes hidden-layer error δ1₁ using h1, (1−h1), and weighted sum of output deltas through W2?

a)

δ1₁ = h1(1−h1)(W2₁₁δ2₁ + W2₂₁δ2₂)

b)

δ1₁ = h1(1−h1)(W2₁₁ + W2₂₁)

c)

δ1₁ = h1(1−h1)(δ2₁ + δ2₂)

d)

δ1₁ = h1(1−h1)(W1₁₁δ2₁ + W1₂₁δ2₂)

88.

After applying learning rate η = 0.6, which update rule is used for W2 elements, for example W2₁,₁?

a)

W2new = W2old − η·δ2·x

b)

W2new = W2old + η·δ2·x

c)

W2new = W2old − η·δ1·h

d)

W2new = W2old + η·δ2·h

89.

Which challenge most directly concerns selecting appropriate learning rate and momentum values in back‑propagation networks?

a)

Selecting a weight update formalism

b)

Choosing suitable non‑linear activation function

c)

Computing error ratio sample‑wise or cumulatively

d)

Difficulty determining parameter values reliably

e)

Deciding hidden layer count and node quantity

90.

What is a typical approach for deciding the reasonable number of hidden nodes in a back‑propagation model when no canonical rule exists?

a)

Conduct targeted experiments iteratively

b)

Use fixed formula from theory

c)

Always match input feature count

d)

Choose maximum nodes possible

e)

Rely on default library settings

91.

Why does the hidden layer size strongly influence convergence in back‑propagation networks?

a)

It fixes learning rate automatically

b)

It guarantees global minima found

c)

It eliminates activation choice

d)

It removes need for bias nodes

e)

It alters weight search complexity

92.

When selecting activation functions for back‑propagation, what is the key consideration stated?

a)

Avoid any non‑linearities

b)

Select based on output size

c)

Use sigmoid exclusively

d)

Prefer linear functions always

e)

Choose one or more carefully

93.

Which statement best captures the issue around weight update approaches in back‑propagation?

a)

Multiple formalisms exist; pick carefully

b)

There is one standard method always

c)

Updates never affect convergence

d)

Weights only change at initialization

e)

Momentum is unnecessary and harmful

94.

What uncertainty surrounds bias nodes in back‑propagation networks?

a)

Whether they store training data

b)

Whether they must be negative

c)

How many to include per layer

d)

If they replace hidden neurons

e)

Why they prevent overfitting

95.

Which methods are contrasted for error ratio computation in back‑propagation training?

a)

Probabilistic versus deterministic

b)

Sample‑by‑sample versus accumulative

c)

Static versus dynamic typing

d)

Supervised versus unsupervised

e)

Analytical versus symbolic math

96.

A practitioner faces no canonical guidance on learning rate η. What is the recommended path forward?

a)

Freeze η at 0.1 for all tasks

b)

Derive η from bias counts

c)

Run experiments to choose values

d)

Use α to replace η entirely

e)

Ignore η and train anyway

97.

In a network with multiple hidden layers, what complication is emphasized?

a)

Eliminating the output layer entirely

b)

Fixing weights without training

c)

Needing only one activation function

d)

Determining nodes per hidden layer

e)

Removing biases from all layers

98.

What pairing of parameters is explicitly cited as difficult to set in back‑propagation?

a)

Kernel size and stride

b)

Input scaling and centering

c)

Dropout rate and seed

d)

Batch size and epochs

e)

Learning rate and momentum

99.

Which choice reflects a realistic misconception about activation functions to test understanding?

a)

Hidden node count determines function

b)

Bias nodes decide the function used

c)

Linear functions suffice for all tasks

d)

Non‑linear options must be considered

e)

Error ratio method selects function

100.

When comparing error computation strategies during training, what must be decided?

a)

Use per‑sample or accumulate errors

b)

Fix hidden nodes to equal inputs

c)

Avoid biases in all layers

d)

Select linear activation always

e)

Prefer momentum over learning rate

101.

Who introduced the neural network known as the Kohonen network?

a)

Geoffrey Hinton

b)

John Hopfield

c)

Teuvo Kohonen

d)

Yann LeCun

102.

Which learning paradigm is primarily used in Kohonen neural networks?

a)

Reinforcement learning

b)

Unsupervised learning

c)

Semi-supervised learning

d)

Supervised learning

103.

What principle drives weight adaptation in Kohonen networks?

a)

Competitive learning

b)

Hebbian supervision

c)

Backpropagation gradients

d)

Bayesian inference

104.

In a typical Kohonen network architecture, how many layers are emphasized?

a)

One processing layer

b)

Two main layers

c)

Three deep layers

d)

Four hierarchical layers

105.

Which metric commonly determines the winning unit for an input in a Kohonen map?

a)

Cosine similarity

b)

Hamming distance

c)

Euclidean distance

d)

Manhattan distance

106.

What characterizes the neighborhood function in a Kohonen map?

a)

Ignores spatial arrangement

b)

Updates only the winner

c)

Updates distant neurons strongly

d)

Updates nearby neurons together

107.

Which application aligns best with Kohonen networks?

a)

Sequence prediction

b)

Image segmentation

c)

Clustering of high-dimensional data

d)

Exact function approximation

108.

What is a typical topology for a Kohonen neural network?

a)

Fully connected recurrent ring

b)

Two-dimensional lattice map

c)

Tree-structured decision nodes

d)

Feedforward multilayer perceptron

109.

In a Kohonen neural network, what does D(j) represent when comparing a sample to cluster j?

a)

Sum of squared feature differences

b)

Dot product between sample and weights

c)

Learning rate scaled distance

d)

Euclidean norm of weight vector

110.

When selecting the winning cluster for a sample in Kohonen learning, which criterion is used?

a)

Minimum D(j) across clusters

b)

Maximum D(j) across clusters

c)

Largest learning rate alpha

d)

Highest input node count

111.

Which step occurs immediately after computing D(j) for all clusters during one sample update?

a)

Increase the sample index

b)

Decrease the number of clusters

c)

Select the minimum distance

d)

Regenerate the weight matrix

112.

What is the standard weight update for the winning cluster column j in Kohonen learning?

a)

w_i,j = w_i,j − αX_i

b)

w_i,j = αX_i + w_i,j

c)

w_i,j = w_i,j + α(X_i − w_i,j)

d)

w_i,j = w_i,j / α

113.

During initialization of a Kohonen network, which parameters are set before training begins?

a)

Kernel size, stride, padding scheme

b)

Input nodes N, clusters M, samples P, α

c)

Bias terms, dropout rate, momentum

d)

Hidden layers, activation type, batch size

114.

Which linear update rule for the learning rate α is valid with β > 0?

a)

α_new = α − β

b)

α_new = α × β

c)

α_new = α + β

d)

α_new = log(α)

115.

Which non-linear alpha update uses a multiplicative decay with 0 < β < 1?

a)

α_new = α − β

b)

α_new = α + β

c)

α_new = α / β

d)

α_new = α β

116.

In the non-linear schemes, which choice increases α when β > 1?

a)

α_new = α − β

b)

α_new = log(α)

c)

α_new = α β

d)

α_new = α / β

117.

In the example, how many clusters are used in the Kohonen network?

a)

Two clusters are defined

b)

Three clusters are defined

c)

Four clusters are defined

d)

One cluster is defined

118.

Which formula is used to compute D(j) for a sample in the worked example?

a)

Sum of squared input–weight differences

b)

Dot product of input and weights

c)

Sum of absolute input–weight differences

d)

Product of inputs and weights

119.

For the first sample 1 1 0 0 0, which cluster wins based on the minimum D(j)?

a)

Cluster j = 1 wins

b)

Cluster j = 2 wins

c)

Both clusters tie

d)

No cluster is selected

120.

After the first sample update with α = 0.6, what is the new weight w1,1?

a)

0.72 after update

b)

0.70 after update

c)

0.68 after update

d)

0.76 after update

121.

For the sample 0 0 0 1 1, which cluster has the smaller distance D(j)?

a)

Decision cannot be determined

b)

Cluster j = 1 has smaller distance

c)

Cluster j = 2 has smaller distance

d)

Distances are equal

122.

When α is updated by α := α − β with α = 0.6 and β = 0.1, what is the new learning rate?

a)

0.6 after reduction

b)

0.7 after reduction

c)

0.4 after reduction

d)

0.5 after reduction

123.

Using α = 0.5, for the first sample re-run, which cluster wins given d1 = 0.9235 and d2 = 2.6544?

a)

Both clusters win equally

b)

No winner due to tie

c)

Cluster 2 wins with smaller d2

d)

Cluster 1 wins with smaller d1

124.

Across iterations, what operation is repeatedly applied to the winning cluster’s weights?

a)

Randomly reset weights to zero

b)

Move weights away from the input vector

c)

Move weights toward the input vector

d)

Swap weights between clusters