Machine Learning

Final Exam
Time Allowed : 2 Hours.
Level Four | June 2022

Choose the most correct answer (2 marks per question):

1. What is true about Machine Learning?

A. Machine Learning (ML) is the field of computer science.
B. ML is a type of artificial intelligence that extract patterns out of raw data by using an
algorithm or method.
C. The main focus of ML is to allow computer systems learn from experience without
being explicitly programmed or human intervention.
D. All of A,B,C
2. Each classifier has a hyper parameter where the learning phase aims to find its finite
optimal value?
A. True.
B. False partially.
C. Partially Correct
D. False.
3. Artificial Neural Networks power lies behind its ability to build and construct the most
curated decision boundary, its main drawback is their bureaucracy or resources
harvesting. The following diagram shows a dummy ANN:
How many hidden layers does this MLP has:

A. 1 B.2 C.3 D.4

4. The weight connection between the input neuron (corresponding to x1) and the 2nd neuron
in 1st hidden layer is:
A. w11 B. w12 C. w21 D. w22
5. For this kind of classifier, the output of each neuron is formulated (Forced) to be in a
specific mathematical range for the sake of regularization effective reproducibility of

1
 ANN. What kind of appropriate type of (CS.AI.ML.Model.ANN.Layer.Output) to bypass
negative values as zeros and else as is:
A. SoftMax B. Narx Sigmoid C. Relu D. Identity
6. The weight connection between the input neuron (corresponding to x2) and the 3rd neuron
in 2nd hidden layer is:
A. w32 B. w23 C. wPhi D. wInf
7. Considering the output layer in the upper ANN figure, We have …. classes.
A. a B. b C. c D. d
8. In every day taxonomy problem we aim to solve using ANN, The output layer
size(Number of neurons) is exact as number of classes:
A. True B. True Partially C. False Partially D. False
9. The hard worker part of ANN lies in the …. layers, because same as the principal of SVM
to do data projection. Those layers are corresponding to other translation and mapping in
order to reduce the redundancy and results stability.
A. Receptive B. Intermediate C. Transfer D. Output
10. In the Linear discriminant classifier LDC,

We use the simplest form of boundary expression in-between the data points holding a
diverse class membership. Simply we goes with the class with:
A. ArgMax{g(x)} B. ArgMin{g(x)} C. ArgMiniMax{g(x)} D. ArgAvrg{g(x)}
11. For LDC, The following equation expression describes…. matrix.

A. Confusion B. Identity C. Covariance D. Variance

12. The SVM classifier is one of the most famous taxonomy algorithms, due to its mighty
ability when needed to project the input patterns from a dimension R Θ →R….
A. -Inf B. Constant C. Target-Output D. Else
13. A machine learning problem involves four attributes plus a class. The attributes have 4, 3,
10, and 2 possible values each. The class has 2 possible values. How many maximum
possible different examples are there?
A.84 B.840 C.48 D.480
14. Machine Learning algorithms usually needs an interfering from us when we use RAW
data, What is the reasonable sequence of steps appropriate to describe this:
A. Normalize data (MiniMax for ex.) → PCA .
B. PCA → normalize PCA output.
C. Normalize data → PCA → Normalize PCA.
D.All A,B,C are possibly True.

2
15. How do you safely handle missing or corrupted data in a dataset?
A. Drop missing rows or columns
B. Impute missing values with mean/median/mode,...etc.
C. Assign a unique category to missing values
D. All of the above
16. The most widely used metrics and tools to assess a classification model are:
A. Confusion matrix
B. Cost-sensitive accuracy
C. Area under the curve (AUC)
D. Receiver Operating Curve (ROC)
17. Suppose we would like to perform clustering on spatial data such as the geometrical
locations of houses. We wish to produce clusters of many different sizes and shapes.
Which of the following methods is the most appropriate to this every kind of data?
A. Fuzzy C-Means.
B. Model-based clustering.
C. K-means clustering.
D. Density-based clustering.
18. Which of the following is a disadvantage of decision trees?
A. Factor analysis
B. Decision trees are robust to outliers
C. Decision trees are prone to be over-fit
D. None of the above
19. I’m the most stable Classifier, No matter number of runs I perform on same data division,
I Guarantee Same Results so I’m:
A. DT B.ANFIS C.K-NN D.Multiple-Regression
20. I’m the most unstable Classifier, For all runs I perform on same data division, I
Guarantee NEW Results so I’m:
A. DT B.ANN C.K-NN D.lINEAR-Regression
21. Which of the following is true about Naive Bayes?
A. Assumes that all the features in a dataset are equally important
B. Assumes that all the features in a dataset are independent
C. Both A and B
D. None of the above options
22. In which of the following cases will K-means clustering fail to give good results?
1. Data points with outliers
2. Data points with different densities
3. Data points with non-convex shapes
A. 1 & 2
B. 1, 2, & 3
C. 2 &
D. 1 & 3

3
23. Which is FALSE regarding regression?
A. It may be used for interpretation
B. It is used for prediction
C. It discovers causal relationships
D. It relates inputs to outputs
24. Real-Time decisions, Game AI, Learning Tasks, Skill Aquisition, and Robot Navigation
are applications of which of the folowing
A. Supervised Learning: Classification
B. Reinforcement Learning
C. Unsupervised Learning: Clustering
D. Unsupervised Learning: Regression
25. Fraud Detection, Image Classification, Diagnostic, and Customer Retention are
applications in which of the following
A. Unsupervised Learning: Regression
B. Supervised Learning: Classification
C. Unsupervised Learning: Clustering
D. Reinforcement Learning

Assume Any Missing Data

4
 Machine Learning
Midterm Exam
Time Allowed : 1 Hours.
Level Four | June 2022

Choose the most correct answer (1 marks per question):

1. Which of the following is a widely used and effective machine learning algorithm based
on the idea of bagging?
A. Decision Tree
B. Random Forest
C. Regression
D. Classification
2. What is perceptron?
A. A single layer feed-forward neural network with pre-processing
B. A neural network that contains feedback
C. A double layer auto-associative neural network
D. An auto-associative neural network
3. Which is true for neural networks?
A. Each node computes it‟s weighted input
B. Node could be in excited state or non-excited state
C. It has set of nodes and connections
D. All of the above
4. Bayes rule can be used for:-
A. Solving queries
B. Increasing complexity
C. Answering probabilistic query
D. Decreasing complexity
5. Naïve Bayes Algorithm is a learning algorithm.
A. Ensemble
B. Reinforcement
C. Unsupervised
D. Else
6. What are the difficulties with k-nearest neighbour algorithm?
A. Calculate the distance of the test case from all training cases
B. Curse of dimensionality
C. Both A & B
D. None of these

5
7. What is/are true about Distance-weighted KNN?
A. The weight of the neighbour is considered
B. The distance of the neighbour is considered
C. Both A & B
D. None of these
8. The points which are at the exact minimum distance corresponding to the minimum
margin in a support vector machine boundaries are called
A.Noise points B.Support Vectors C- Product Vectors D) none of the
above
9. In order to project the original space into a new high-dimensional space Support Vector
Machine uses the
A. Kernel trick B. Optimal Linear boundary C. Cumulative distribution D.
Else
10. Consider the figure that show a trained Support vector machine classifier for a simple
data set the points which are the exact minimum distance corresponding to the min
margin are
A. Noise points B. Support vectors C. Product vectors D. E
11. How the decision tree reaches its decision ?
A. Single test b. Two Test c. Sequence of tests d). zero of
tests
12. The minimum time complexity for training an SVM is O(N 2). According to this fact, what
sizes of datasets are not best suited for Support Vector Machine?
A) Large datasets B) Small datasets C) Medium datasets D)Size does not
matter
13. If I am using all features of my dataset and I achieve 100% accuracy on my
VALIDATION set, but ~70% on TRAINING set, what should I look out for?
A)Under-fitting B)Nothing, the model is perfect C)Over-fitting D)Try another
algorithm
14. The maximum possible accuracy on a test set may vary from classification algorithm to
another , Sometimes we never reach a 100% accuracy due to the lack of the every possible
hyper-parameters that accomplish that. In such scenarios we should find a candidate
solution such as:
A. Data Manipulation. B. Outlier Removal.
C. use other representation. D. Label Considered as Annotations.
15. Clustering and classifications are such a twin of algorithms, one of them produce a
description for the data (Annotation) from scratch, meanwhile classification can be used
to prove the validity of this LABELING process. Both utilized in the US. Dollar crisis to
regroup bank customers and generate new profiles (Labels) for them. Clustering is always
useful no matter we have the data TRUE labels. But classification mainly need Labels.
A. True B. True Partially C. False Partially D. False

6
Assume Any Missing Data