Fathima Jamal-Deen 

Dashboard  My courses  5CCS2INT 21~22 SEM2 000001 ARTIFICIAL INTE  Week 1: Introduction to Artificial
Intelligence & The Principles of Search  Quiz 1

Started on Wednesday, 18 May 2022, 4:52 PM

State Finished
Completed on Wednesday, 18 May 2022, 7:07 PM
Time taken 2 hours 14 mins
Grade 10 out of 10 (100%)

Question 1 Correct Mark 2 out of 2

Which of the following are valid properties of an AI agent?

A. Efficient: It makes decisions as fast as possible.

B. Observant: It completely understands every facet of a problem such that it can act upon it.

C. Autonomous: It makes decisions without human input

D. Rational: It makes decisions based on the knowledge it has.

E. Responsive: It reacts to changes happening in the world around it.

Question 2 Correct Mark 2 out of 2

Which of the following are components of the state transition system Σ?

A. Set of all actions A

B. Set of all states S

C. State transition function γ

D. Successor state s'

E. Current action a
Question 3 Correct Mark 2 out of 2

When running either approach to uninformed search, what is the overall structure of the
algorithm?
Assume that you are using both an open and closed list for this question.  

1. While open list is not empty

2. Assign current state as next state in the open list

3. If current state is the goal: end search

4. Else: add current state to closed list

5. Expand current state to check for successors

6. Loop through all successors

7. If successor is not in closed list: add to open list

Question 4 Correct Mark 2 out of 2

Why is it important that for uniform cost search and A* that g(x) counts the total cost from the
initial state?  Select all that apply.

A. The action for a given state may have a low cost, but the path to reach it may be longer than
others.

B. We cannot determine how expensive an action is without the costs that preceded it.

C. We may visit the same state twice from different paths and need to know the cost taken to
reach it.

D. The goal state needs to factor the total distance reached.

E. The total cost will help us recognise whether the current state is closer to the goal.
Question 5 Correct Mark 2 out of 2

Which of the following properties of heuristics is correct?  Select all that apply.

a. A heuristic that always estimates the distance to the goal as more than the actual cost
is admissible.

b. Heuristics are always easy to calculate.

c. A heuristic with a value of 0 is valid.

d. A heuristic may be calculated from a relaxation of the problem.

e. A heuristic that always estimates the distance to the goal as less than or equal to the actual cost
is admissible.

◄ Tutorial Examples: Search

Jump to...
Adversarial Search 01: Minimax ►
 Fathima Jamal-Deen 
Dashboard  My courses  5CCS2INT 21~22 SEM2 000001 ARTIFICIAL INTE  Week 2: Adversarial Search &
Introduction to CSPs  Quiz 2

Started on Wednesday, 18 May 2022, 5:03 PM

State Finished
Completed on Wednesday, 18 May 2022, 5:05 PM
Time taken 2 mins 4 secs
Grade 10.00 out of 10.00 (100%)

Question 1 Correct Mark 1.00 out of 1.00

Which of the following properties are true of an adversarial search such as Minimax?  Select all
that apply.

A. You can safely ignore the behaviour of other agents that are not your own.

B. There is more than one agent trying to solve problems in the same search space.

C. If there are multiple agents in the search space, their goals contradict and conflict with one
another.

D. Exploration through the state space will alternate between active agents.

E. If there are multiple agents in the search space, their goals align and do not conflict with one
another.
Question 2 Correct Mark 3.00 out of 3.00

Consider the diagram below, what is the utility calculated for this tree using Minimax?  Assume
that the first layer from the root is calculating the MAX player.

Answer: 5

Question 3 Correct Mark 2.00 out of 2.00

Why is Alpha-Beta pruning a useful technique to employ in Minimax?  Select all that apply.

A. We can find better solutions for the MAX player.

B. State spaces can become so large that searching with Minimax can take a long time.

C. We can remove the need to run Minimax on future nodes.

D. It helps ensure our heuristics will stay admissible.

E. We can determine quickly whether a specific subtree is going to be worth exploring based on
the current values of alpha and beta.
Question 4 Correct Mark 2.00 out of 2.00

What are the three key components of a constraint satisfaction problem?

A. A set of constraints that dictates relationships between domains.

B. A set of constraints that dictates relationships between variables.

C. A finite series of variables.

D. A set of domains for constraints to select variables.

E. A set of domains for variables to be assigned from.

Question 5 Correct Mark 2.00 out of 2.00

Which of the following approaches are valid for solving a CSP?  Select all that apply.

A. Forward checking to identify constraint violations early.

B. Assigning variables that appear to be the most constrained.

C. Forward checking to fix constraint violations early.

D. Assigning values to variables that appear to be the least constraining on others.

E. Assigning variables that appear to be the least constrained.

◄ Lecture Slides - Constraint Satisfaction [PPTX]

Jump to...
Tutorial Example: CSPs ►
 Fathima Jamal-Deen 
Dashboard  My courses  5CCS2INT 21~22 SEM2 000001 ARTIFICIAL INTE  Week 3: Introduction to Automated
Planning  Quiz 3

Started on Wednesday, 18 May 2022, 5:15 PM

State Finished
Completed on Wednesday, 18 May 2022, 5:17 PM
Time taken 1 min 40 secs
Grade 10.00 out of 10.00 (100%)

Question 1 Correct Mark 2.00 out of 2.00

Which of the following are true of relations in a Constraint Satisfaction Problem?  Select all that
apply.

A. The constraint arity k is based on a relation between a subset of k domains

B. If our domain has two values to select from, it's a binary relation.

C. If we only have two variables in a problem, it's a binary relation.

D. The Australian map colour problem has tertiary relations.

E. Binary relations are expressed across pairs of variables.

Question 2 Correct Mark 2.00 out of 2.00

When we are trying to solve problems through planning, what are we trying to figure out?  Select
all that apply.

A. Select which actions we should execute.

B. How are actions executed in the problem.

C. Why actions should be executed in the problem.

D. What actions exist in the problem.

E. In what order should we execute the actions we have selected.

Question 3 Correct Mark 2.00 out of 2.00

What does PDDL stand for?

Answer: planning domain definition language

Question 4 Correct Mark 2.00 out of 2.00

Which of the following are components we will expect to find in a PDDL action?  Select all that
apply.

A. The effects of applying the action.

B. The preconditions that need to hold true to execute the action.

C. Parameters used for the action.

D. The states that this action can be applied in.

E. The name of the action.

Question 5 Correct Mark 2.00 out of 2.00

If we consider the Gripper problem from the slides, which parts of a given problem appear in
which part of the PDDL?  Select all that are true. 

A. The action definitions are in the problem file.

B. The action definitions are in the domain file.

C. The individual boxes, grippers and rooms for a given problem are defined in the domain file.

D. The individual boxes, grippers and rooms for a given problem are defined in the problem file.

E. The definition of boxes, grippers and rooms are defined in the domain file.

◄ Tutorial Example: PDDL

Jump to...
Planning in PDDL 01: Metric Planning ►
 Fathima Jamal-Deen 
Dashboard  My courses  5CCS2INT 21~22 SEM2 000001 ARTIFICIAL INTE  Week 4: Expanding on PDDL and
Relaxed Planning  Quiz 4

Started on Wednesday, 18 May 2022, 7:14 PM

State Finished
Completed on Wednesday, 18 May 2022, 7:15 PM
Time taken 48 secs
Grade 10.00 out of 10.00 (100%)

Question 1 Correct Mark 2.00 out of 2.00

How do we enable the use of metrics and numbers in PDDL 2.1?

Select one or more:

A. You add the fluents requirement in the domain file.

B. You add the fluents requirement in the problem file.

C. You add the metrics requirement in the domain file.

D. You add the metrics requirement in the problem file.

E. Trick question: none of these are correct.

Question 2 Correct Mark 2.00 out of 2.00

Which of the following terms can be applied to the condition of a PDDL 2.1 durative action? 
Select all that apply.

A. 'over all' that tracks facts that must be true throughout the complete duration of the action

B. 'duration' that tracks how long the conditions must hold for

C. 'at end' that tracks facts that must be true at the end of the action

D. 'at start' that tracks facts that must be true at the start of the action
Question 3 Correct Mark 2.00 out of 2.00

What are domain-independent heuristics used in AI planners? Select all that apply.

A. A heuristic that can be calculated irrespective of the information in the PDDL domain file. 

B. A guaranteed admissible heuristic.

C. A heuristic that can be calculated without using specific information from the PDDL problem
file. 

D. A heuristic calculated by analysing the structure of search space itself, rather than the PDDL
facts.

E. A heuristic that can solve any problem.

Question 4 Correct Mark 2.00 out of 2.00

Why do AI Planners require the use of domain-independent heuristics? Select all that apply.

A. Users cannot enter a heuristic into the PDDL domain file.

B. A planner does not need a heuristic to search effectively.

C. Planners do not analyse domain/problem files to create a bespoke heuristic.

D. A planner should be able to receive any valid PDDL domain/problem and solve it without a
heuristic being provided.

E. Users cannot enter a heuristic in the PDDL problem file.

Question 5 Correct Mark 2.00 out of 2.00

Which of the following are valid behaviour of the FF planner?  Select all that apply.

A. Best-First Search is applied if no valid successor while running EHC

B. States are assessed through the use of the RPG heuristic.

C. Best-First Search is applied if the EHC fails.

D. All successors (children) of a state are assessed in the Enforced Hill Climber 

E. Breadth-First Search is applied if the EHC fails.

◄ Satellite RPG
Jump to...
Tutorial Example: RPG Planning ►
 Fathima Jamal-Deen 
Dashboard  My courses  5CCS2INT 21~22 SEM2 000001 ARTIFICIAL INTE  Week 5: Handling Uncertainty
 Quiz 5

Started on Wednesday, 18 May 2022, 7:11 PM

State Finished
Completed on Wednesday, 18 May 2022, 7:12 PM
Time taken 1 min
Grade 10.00 out of 10.00 (100%)

Question 1 Correct Mark 1.00 out of 1.00

How might our certainty in a problem come into question?  Select all that apply.

A. Our actions might have unintended outcomes.

B. We might not have all the relevant information about the current state that we need to act. 

C. The behaviour of other systems/processes may prove random and difficult to understand.

D. There could be ethical implications we have not yet considered in the work.

E. The problem we're trying to solve might not be the one we want to be solving.

Question 2 Correct Mark 1.00 out of 1.00

Which of the following situations are examples of a non-deterministic system?

A. Tossing a coin.

B. Playing a Fruit/Slot machine.

C. Playing a game of Texas Hold'em Poker.

D. Playing a game of Tic-Tac-Toe.

E. Using a random number generator.

Question 3 Correct Mark 1.00 out of 1.00

Which of the following situations are examples of a partially-observable system?

A. Playing a game of Solitaire.

B. Playing a game of Chess.

C. Playing a game of Texas Hold'em Poker.

D. Playing a game of Uno.

E. Playing a game of Tic-Tac-Toe.

Question 4 Correct Mark 2.00 out of 2.00

Which of the following are important properties of a markov chain?  Select all that apply.

A. Probability distributions for state transitions hold and do not change.

B. All states need transitions that go back into themselves.

C. We label all actions clearly.

D. Transitions between states are only influenced by information in the current state.

E. Transitions are given probabilities of possible success.

Question 5 Correct Mark 1.00 out of 1.00

Why are the reward values relevant when we're building policies with MDPs?

Select one or more:

A. Rewards are merely cosmetic and have no influence on decision making.

B. Reward values help give incentive to an agent to make the best decisions possible. 

C. Rewards help remind humans which actions are important.

D. Rewards tell us the exact long-term potential of a given state.

E. Rewards tell the agent if the previous action was the best decision to make.
Question 6 Correct Mark 2.00 out of 2.00

Why would we use negative rewards when designing a MDP?

Select one or more:

A. It doesn't matter if rewards are positive or negative.

B. Negative rewards help clearly denote 'bad' areas of the state space we don't want to visit.

C. Negative rewards are used to make the AI feel bad about itself.

D. Negative rewards incentivise an agent to devise an optimal solution that minimises the overall
penalty. 

E. Negative rewards help remind us which actions are 'bad' actions.

Question 7 Correct Mark 2.00 out of 2.00

What is the purpose of reducing the discount factor γ to be less than 1 when calculating the
Utility of a given state?  Select all that apply.

A. Ensure that the most immediate reward we receive is considered the most relevant when
updating the Utility.

B. To punish the agent for making bad choices later on.

C. Ensure that future rewards are still relevant, but not as relevant as the most immediate reward.

D. To help the Utility values develop a value more reflective of the potential of the current state.

E. To make sure the Utility values don't converge too quickly.

◄ Lecture Slides (Handling Uncertainty) [PPTX]

Jump to...
Tutorial Exercises: Uncertainty ►
 Fathima Jamal-Deen 
Dashboard  My courses  5CCS2INT 21~22 SEM2 000001 ARTIFICIAL INTE  Week 6: Machine Learning (I):
Unsupervised Learning  Quiz 6

Started on Wednesday, 18 May 2022, 7:26 PM

State Finished
Completed on Wednesday, 18 May 2022, 7:27 PM
Time taken 54 secs
Grade 10.00 out of 10.00 (100%)

Question 1 Correct Mark 2.00 out of 2.00

Clustering is considered to be what kind of machine learning?

A. Reinforcement Learning

B. Deep Learning

C. Supervised Learning

D. Associative Learning

E. Unsupervised Learning

Question 2 Correct Mark 2.00 out of 2.00

In the context of K-Means Clustering, what does the K represent?  Select all that apply.

A. The maximum number of clusters we need to maintain (but we can have less than this if
necessary).

B. The minimum number of clusters we need to maintain (but we can have more than this if
necessary).

C. The total number of clusters we will maintain throughout the process.

D. The level of the hierarchy we wish to cut from.

E. The number of initial clusters the process starts with.

Question 3 Correct Mark 2.00 out of 2.00

Which of the following facts are true of K-Means Clustering?  Select all that apply.

A. Cluster positions/distances are influenced by the points that enter them.

B. Once a point is assigned to a cluster, it cannot move to a different one.

C. Points are assigned to clusters based on their distance to each centroid.

D. Clusters are immutable and never change.

E. New cluster centroids are calculated as the average of all points assigned to that cluster.

Question 4 Correct Mark 2.00 out of 2.00

Which of the following facts are true of Hierarchical Clustering?  Select all that apply.

A. New cluster centroids are calculated as the average of all points assigned to that cluster.

B. Once a point is assigned to a cluster, it cannot move to a different one.

C. Clusters are immutable and never change.

D. Cluster positions/distances are influenced by the points that enter them.

E. Points are assigned to clusters based on their distance to one another.

Question 5 Correct Mark 2.00 out of 2.00

What is the benefit of cutting the dendrogram at different levels in Hierarchical Clustering?  Select
all that apply. 

A. The individual values can be manually moved into new clusters quickly.

B. Clusters stay the same, but the complexity of their data changes.

C. We can reset the K value by cutting.

D. We can manually decide on useful clusters by looking at the dendrogram.

E. We can create different sets of clusters from the same dendrogram by cutting at different
levels.

◄ Lectures Slides (Unsupervised Learning) [PPTX]

Jump to...
Tutorial Examples: Unsupervised Learning ►
 Fathima Jamal-Deen 
Dashboard  My courses  5CCS2INT 21~22 SEM2 000001 ARTIFICIAL INTE  Week 7: Machine Learning (II):
Supervised Learning  Quiz 7

Started on Wednesday, 18 May 2022, 7:36 PM

State Finished
Completed on Wednesday, 18 May 2022, 7:36 PM
Time taken 37 secs
Grade 10.00 out of 10.00 (100%)

Question 1 Correct Mark 2.00 out of 2.00

Which of the following could be considered a supervised learning classification problem?  Select
all that apply.

A. Recognising someone's handwriting to translate into a digital document.

B. Identifying toxic comments on social media.

C. Detecting a specific object in a photograph.

D. The Mars rover moving to collect a rock sample.

E. Estimating cost of materials for importing.

Question 2 Correct Mark 2.00 out of 2.00

Which of the following facts are true of Neural Networks?  Select all that apply.

A. Neurons fire provided they receive a non-zero input.

B. You must have one or more hidden layers in a neural network.

C. The TanH transfer function will ensure prevent negative outputs.

D. Neuron input is a weighted-summation function.

E. Network weights can be trained via back-propagation.

Question 3 Correct Mark 2.00 out of 2.00

Which of the following is not a valid regression technique?  Select all that apply.

A. Linear Regression

B. K-Means

C. Neural Networks

D. Decision Trees

E. Polynomial Regression

Question 4 Correct Mark 2.00 out of 2.00

Consider the following image and select all facts that apply.

A. This data has zero correlation between the X and Y variables.

B. This data has a positive correlation between the X and Y variables.

C. This data has a weak correlation between the X and Y variables.

D. This data has a negative correlation between the X and Y variables.

E. This data has a strong correlation between the X and Y variables.

Question 5 Correct Mark 2.00 out of 2.00

Consider the following image and select all facts that apply.

A. The regression line has a slope coefficient of around 2.1

B. The regression line has an intercept of approximately 0.95

C. The regression line has an intercept of approximately 2.1

D. This regression line has a high error rate (SSE).

E. The regression line has a slope coefficient of around 0.95

◄ Lecture Slides (Supervised Learning) [PPTX]

Jump to...
Tutorial Examples: Supervised Learning ►
 Fathima Jamal-Deen 
Dashboard  My courses  5CCS2INT 21~22 SEM2 000001 ARTIFICIAL INTE  Week 8: Machine Learning (III):
Reinforcement Learning  Quiz 8

Started on Wednesday, 18 May 2022, 7:50 PM

State Finished
Completed on Wednesday, 18 May 2022, 7:50 PM
Time taken 55 secs
Grade 10.00 out of 10.00 (100%)

Question 1 Correct Mark 1.00 out of 1.00

When is reinforcement learning more pragmatic to use than trying to capture a problem using a
Markov Decision Process (MDP)?  Select all that apply.

A. We are now aware of the complete set of all states.

B. The probability distribution of actions is unknown.

C. The utility of terminal states are not the same.

D. The reward values for R(s) are not the same in each state.

E. The reward values for R(s) are not known for each state.

Question 2 Correct Mark 1.00 out of 1.00

Which of the following are reflective of the behaviour of model-free reinforcement learning
algorithms?  Select all that apply.

A. Generate rollouts and sample returns from state-action pairs.

B. Assume no knowledge of the underlying model of the problem.

C. Estimate the utility function U for all states.

D. The probability distribution of actions is unknown.

E. Converge on the heuristic value as numbers of visits per rollout increases.

Information

For the next two questions, consider these Monte Carlo rollouts.

Question 3 Correct Mark 2.00 out of 2.00

What is V(A) if using First Visit Monte Carlo?  Give your answer up to 2 decimal places.

Answer: -5.67

Question 4 Correct Mark 2.00 out of 2.00

What is V(A) if using Every Visit Monte Carlo?  Give your answer up to 2 decimal places.

Answer: -4.80

Question 5 Correct Mark 2.00 out of 2.00

What is V(B) if using First Visit Monte Carlo?  Give your answer up to 2 decimal places.

Answer: -9.67

Question 6 Correct Mark 2.00 out of 2.00

What is V(B) if using Every Visit Monte Carlo?  Give your answer up to 2 decimal places.

Answer: -8.80
 ◄ Lecture Slides (Reinforcement Learning) [PPTX]

Jump to...
Reinforcement Learning Tutorial Questions ►
 Fathima Jamal-Deen 
Dashboard  My courses  5CCS2INT 21~22 SEM2 000001 ARTIFICIAL INTE  Week 9: The Importance of Ethical
AI  Quiz 9

Started on Wednesday, 18 May 2022, 7:57 PM

State Finished
Completed on Wednesday, 18 May 2022, 7:57 PM
Time taken 34 secs
Grade 10.00 out of 10.00 (100%)

Question 1 Correct Mark 2.50 out of 2.50

What are the three main areas of consideration for ethical issues in AI?

A. AI that achieves general intelligence

B. AI that is self-aware

C. Applications of AI in the real world.

D. AI that reflects human values or bias.

E. Access and control of AI technologies.

Question 2 Correct Mark 2.50 out of 2.50

Which of the following application areas for AI could present an ethical issue?  Select all that
apply.

A. Image recognition of CCTV footage

B. Supervised learning for medical diagnosis

C. Navigation on road networks to requested destinations

D. AI planning for military strategy

E. Voice recognition in a smart home assistant.


Question 3 Correct Mark 2.50 out of 2.50

Which of the following is a concern when we consider access to AI technologies?

A. AI tools only being driven by business considerations.

B. Machine learning models being trained against limited and hidden datasets.

C. Challenges that could be tackled in minority communities are ignored by AI companies. 

D. AI solutions being so expensive they're not accessible to companies/governments in poorer

economies.

E. All of the above.

Question 4 Correct Mark 2.50 out of 2.50

Why do we need to focus on ensuring human values are reflected in AI systems?  Select all that
apply.

A. AI systems often find solutions that are completely different from our expectations.

B. AI systems may find solutions that enable self-awareness.

C. AI systems might inadvertently reflect human behaviour which is undesirable.

D. AI can often find cheap and effective solutions that may be immoral or unethical.

E. AI systems is not a concern: it is perfectly reliable as a mechanism to search for solutions.

◄ Ethical AI Slides

Jump to...
‘Self-driving’ cars are still a long way off. [The Conversation] ►