test1

Question 1

1. What is the primary goal of supervised learning?
   A) To find hidden patterns in data
   B) To predict outcomes based on labeled examples
   C) To group data into clusters
   D) To generate completely new data

Answer 1

B) To predict outcomes based on labeled examples

Question 2

2. What kind of problem is the “Punches-kicks-fall” problem?
   A) Regression
   B) Classification
   C) Clustering
   D) Dimensionality Reduction

Answer 2

B) Classification

Question 3

3. Which of the following is NOT a component of a perceptron?
   A) Inputs
   B) Weights
   C) Bias
   D) Hidden Layers

Answer 3

D) Hidden Layers

Question 4

4. If the perceptron learning algorithm finds a solution, what does this indicate about the problem?
   A) It is linearly separable
   B) It is non-linearly separable
   C) It has no possible solution
   D) The algorithm is incorrect

Answer 4

A) It is linearly separable

Question 5

5. What does the perceptron training rule do when an error is detected?
   A) Ignores the error
   B) Updates weights using the error value
   C) Adds random noise to the input
   D) Decreases the learning rate

Answer 5

B) Updates weights using the error value

Question 6

6. What is the main limitation of the perceptron?
   A) It can only classify numerical data
   B) It cannot solve non-linearly separable problems
   C) It requires a lot of computational power
   D) It does not work with large datasets

Answer 6

B) It cannot solve non-linearly separable problems

Question 7

7. What is the advantage of using more hidden layers in a neural network?
   A) It reduces the number of computations
   B) It allows solving more complex problems
   C) It makes training faster
   D) It prevents overfitting

Answer 7

B) It allows solving more complex problems

Question 8

8. Which activation function is commonly used in perceptrons?
   A) Sigmoid
   B) ReLU
   C) Hard threshold (step function)
   D) Softmax

Answer 8

C) Hard threshold (step function)

Question 9

9. What is the key difference between the Perceptron and Adaline?
   A) Adaline uses a different weight update rule based on error minimization
   B) Perceptron uses sigmoid activation
   C) Adaline cannot be trained
   D) Perceptron uses gradient descent

Answer 9

A) Adaline uses a different weight update rule based on error minimization

Question 10

0. What is the formula for weight update in Adaline?
   A)
   𝑤
   =
   𝑤
   +
   𝛼
   𝑥
   w=w+αx
   B)
   𝑤
   =
   𝑤
   +
   𝛼
   𝑥
   𝐸
   w=w+αxE
   C)
   𝑤
   =
   𝑤
   +
   𝑥
   𝐸
   w=w+xE
   D)
   𝑤
   =
   𝑤
   −
   𝛼
   𝑥
   w=w−αx

Answer 10

B)
𝑤
=
𝑤
+
𝛼
𝑥
𝐸
w=w+αxE

Question 11

11. What is the function of the learning rate (α) in weight updates?
    A) It determines the speed of convergence
    B) It increases the complexity of the model
    C) It removes noise from the data
    D) It changes the input values

Answer 11

A) It determines the speed of convergence

Question 12

12. Which type of neural network uses convolution operations for feature extraction?
    A) Recurrent Neural Networks (RNNs)
    B) Convolutional Neural Networks (CNNs)
    C) Multi-layer Perceptrons (MLPs)
    D) Support Vector Machines (SVMs)

Answer 12

B) Convolutional Neural Networks (CNNs)

Question 13

13. What is the gradient in gradient descent?
    A) The sum of all errors
    B) The vector of partial derivatives of the error function
    C) The final weight values after training
    D) A constant value used for optimization

Answer 13

B) The vector of partial derivatives of the error function

Question 14

14. Why does the perceptron fail to classify XOR correctly?
    A) The XOR function is not linearly separable
    B) XOR has too many inputs
    C) The perceptron cannot handle binary inputs
    D) The learning rate is too high

Answer 14

A) The XOR function is not linearly separable

Question 15

15. What is backpropagation used for?
    A) To initialize weights randomly
    B) To update weights in multi-layer neural networks
    C) To generate new input data
    D) To decrease the number of layers in a network

Answer 15

B) To update weights in multi-layer neural networks

Question 16

16. What is the difference between a Multi-Layer Perceptron (MLP) and a Deep Neural Network (DNN)?
    A) MLPs have more than one hidden layer, while DNNs only have one
    B) DNNs have multiple hidden layers, while MLPs typically have one
    C) DNNs only perform classification, while MLPs perform regression
    D) There is no difference; they are the same

Answer 16

B) DNNs have multiple hidden layers, while MLPs typically have one

Question 17

17. What does increasing the number of training examples generally do?
    A) Reduces overfitting
    B) Decreases accuracy
    C) Makes training slower with no benefit
    D) Eliminates the need for a learning rate

Answer 17

A) Reduces overfitting

Question 18

18. Which function is commonly used for regression problems in neural networks?
    A) Softmax
    B) ReLU
    C) Sigmoid
    D) Linear activation

Answer 18

D) Linear activation

Question 19

19. What does the sigmoid activation function do?
    A) It transforms values into probabilities between 0 and 1
    B) It outputs only binary values
    C) It prevents gradient descent from working
    D) It makes training slower

Answer 19

A) It transforms values into probabilities between 0 and 1

Question 20

20. What is the primary purpose of using multiple layers in neural networks?
    A) To store more data
    B) To reduce computation time
    C) To learn complex representations of data
    D) To make the network easier to interpret

Answer 20

C) To learn complex representations of data