Question

With the aid of a diagram, discuss the Inference engine cycle via a match fire procedure.

Question
Explain the concept of Probabilistic Reasoning.

Question
Pinku has installed a new burglary alarm. It is reliable to detect a burglary, but it also responds to
minor earthquakes. Consider Pinku has two neighbours, Akshat and Aakash who promised Pinku to
call him at work when they hear the alarm. Akshat nearly always calls when any alarm is heard. But
sometimes he gets confused with the telephone ringing and the alarm, and calls then, too. Aakash likes
loud music and often misses the alarm altogether. In the below figure, we can identify that who has
called and who has not.

a. Estimate the probability of a burglary.

b. Calculate the probability that the alarm has sounded, but neither an earthquake or burglary occurred, and
both Akshat and Aakash made a call to Pinku.

Question
Describe the basis of Bayesian network specification
Question
Describe how a Turing Test is performed

Question
Discuss with the aid of a diagram the concept of convolutional Neural Networks.

Question
Differentiate between supervised, unsupervised and reinforcement learning and give examples of
each.

Question
What are the main components of a rule–based system?

Question
Describe the phases of designing an expert system? What term is used to call the whole process?

Question
What is the main difference between conventional computer programs and production systems (rule–
based systems)?

Question
a. Using the following data develop a linear regression model for the following data. Show all steps
you followed to create the model including the plot diagram.

Student Marks (%) Attendance (%)

70 89
45 60
24 10
90 50
80 70
12 50
45 60
89 100
23 45
46 90
75 50
100 100
90 100
34 96
10 10
67 68

b. Use the model above to predict the marks for students with an attendance record of 65%.

Question
For the following agents, briefly state possible PEAS (Performance measure, Environment, Actuators,
Sensors).
a. Automated taxi driver

b. Medical diagnostic system

Question
Providing reasons in each case, identify the type of agent depicted by the following diagrams.
Question
The following is the rule set of a simple weather forecast expert system:
1. IF cyclone THEN clouds
2. IF anticyclone THEN clear sky
3. IF pressure is low THEN cyclone
4. IF pressure is high THEN anticyclone
5. IF arrow is down THEN pressure is low
6. IF arrow is up THEN pressure is high

Use forward chaining to reason about the weather if the working memory contains the fact: arrow is
down. Show your answer in a table naming the rules matching the current working memory (conflict
set), which rule you apply, and how the working memory contents changes on the next cycle after a
rule has fired

cycle Working Conflict set Rule fired

Memory

Question
Discuss an appropriate situation to use a single layer perceptron algorithm.

Question
The following is the rule set of a simple weather forecast expert system:
1. IF green THEN walk
2. IF red THEN wait
3. IF green AND blinking THEN hurry
4. IF red OR green THEN traffic light works
Which of the above rules will be put into a conflict set by the system if the working memory contains
two facts: green, blinking? Explain why each rule is selected or not.

Question
Describe at least FOUR (4) advantages that expert systems offer organizations that would otherwise
have to employ human experts.

Question
Use the following diagram to simulate 1 epoch of the error back propagation (EBP) algorithm where
hidden nodes are activated via the sigmoid function. You are also requested to set initial weights
randomly, and other inputs will be initialized as follows:

𝑥1 = 2, 𝑥2 = −1, 𝜃4 = −0.21, 𝜃5 = 0.5, 𝜃6 = 0.25, 𝜂 (𝑙𝑒𝑎𝑟𝑛𝑖𝑛𝑔 𝑟𝑎𝑡𝑒) = 0.02 𝑎𝑛𝑑 𝑦 =1

a. Show (by calculation) how all hidden nodes will be activated. [9 marks]
b. Illustrate (by calculation) how to perform weights training. [12 marks]

Question
Use the binary image below to answer questions a, b, and c.
a. Apply the following 4X4 filters to the image above to detect features and describe the feature map
produced. [Hint: dot product]

1 -1 0 -1
0 1 0 0
0 -1 1 0
-1 -1 0 1

0 -1 2 -1
4 0 0 4
2 2 0 0
-1 -2 0 0

-1 -1 0 -1
0 -1 0 2
2 -1 -1 2
2 -1 0 -1

b. Create a 2D-max-pooling operation for each feature map produced.

c. Illustrate how you would feed the feature images into a neural network [Hint flattening].

Question
Using the Bayesian network below label, the following statements about conditional independence as
true or false. For this question, you should consider only the structure of the Bayesian network, not the
specific probabilities. Explain each of your answers.

i. t and u are independent.

ii. t and u are conditionally independent given i, e, and h.
iii. t and u are conditionally independent given i and h.
iv. e and h are conditionally independent given u.
v. e and h are conditionally independent given u, i, and t.
vi. i and h are conditionally independent given e.
vii. i and h are conditionally independent given t.
viii. t and h are independent.
ix. t and h are conditionally independent given e.
x. t and h are conditionally independent given e and u.

Question
How do you calculate odds ratios in logistic regression?

Question
Differentiate between odds ratios and probability ratios

Question
How is a learning task defined in machine learning?

Question
Engaged in a debate regarding reverse psychology and self-fulfilling prophecy, the IS students decide
to write a Rule-Based system to determine, once and for all, whether Neo should be able to determine
if he is “The One” in the original film “The Matrix”.

They came up with the following rules:

Rule 0: IF (OR '(?x) is a noob', '(?x) is disenchanted')

THEN '(?x) visits The Oracle'

Rule 1: IF '(?x) visits The Oracle'

THEN '(?x) picks The One'

Rule 2: IF (AND '(?x) is a noob', '(?x) visits The Oracle')

THEN 'The Oracle says (?x) is not The One', '(?x) is disenchanted'

Rule 3: IF (AND (OR '(?x) is disenchanted', '(?x) is a noob'),

'The Oracle says (?x) is not The One')
THEN '(?x) practices Kung Fu'

Rule 4: IF (AND '(?x) is disenchanted','(?x) loves (?y)')

THEN '(?x) proclaims love for (?y)')

Rule 5: IF (AND 'The Oracle says (?x) is not The One', '(?y) proclaims love for (?x)'
'(?y) picks The One')
THEN '(?x) is The One'
The following assertions were determined:
Assertion 0: (Neo is a noob)
Assertion 1: (Trinity is disenchanted)
Assertion 2: (Trinity loves Neo)
Assertion 3: (Trinity proclaims love for Neo)

Additionally, the following assumptions about backwards chaining were determined:

 The backward chainer tries to find a matching assertion in the list of assertions. If no matching
assertion is found, the backward chainer tries to find a rule with a matching consequent. In
case none are found, then the backward chainer assumes the hypothesis is false.
 The backward chainer never alters the list of assertions, so it can derive the same result
multiple times.
 Rules and antecedents are tried in the order they appear.

a. Using the Hypothesis that “NEO IS THE ONE”, simulate backward chaining.

i. Write all the hypotheses the backward chainer looks for in the database in chronological order.

ii. Draw the goal tree that would be created by backward chaining from this hypothesis.

Question
List and explain the different types of AI agents

Question
Perform a single step perceptron training for Multi-layer perceptron using Error Back Propagation

Question
You have a friend who only does one of four things on every Saturday afternoon: go shopping, watch
a movie, play tennis, or just stay in. You have observed your friend’s behaviour over 11 different
weekends. On each of these weekends you have noted the weather (sunny, windy, or rainy), whether
her parents visit (visit or no-visit), whether she has drawn cash from an ATM machine (rich or poor),
and whether she had an exam during the coming week (exam or no-exam).

You have built the following data table:

You now want to build a decision tree to predict the activity of your friend on any future Saturday
afternoon from the observed values of Weather, Parents, Cash, and Exam.

a. Suppose that you want to build a very simple decision tree that allows you to predict the value of
Decision from a single observable attribute (Weather, Parents, Cash, or Exam).
What would be this attribute? If there is a tie among several attributes give each of them. For each
attribute that you have selected for each value of this attribute, give the answer that the decision tree
would give.

b. Draw the full tree that correctly classifies all the examples