COMP3308/3608, Lecture 13b
ARTIFICIAL INTELLIGENCE
Preparation for the Exam
Sample exam/revision questions are available on Canvas
Copyright By PowCoder代写 加微信 powcoder
Three documents:
1) Sample exam questions
2) Search: Weeks_2-3_Practice.pdf
3) Bayesian networks: BN practice questions
, COMP3308/3608 AI, week 13b, 2022 1
Marking – Reminder
No more homeworks and assignments – well done for completing all assessments!
Homeworks – the 4 selected homeworks were announced on Ed, marking is completed and the marks are in Canvas
Assignment 2
Code (/12) in Grok – your mark is already available in Grok; it is
determined by the number of tests you have passed
The code mark will be uploaded from Grok to Canvas soon – we are waiting for Grok to fix some issues. You will be asked to check your mark in Canvas.
Report (/12) in Canvas – we aim to complete the marking by Sunday w13
We mark progressively and as soon as your report is marked, you can see your mark in Canvas, together with the breakup of your marks against the marking criteria and additional feedback on your assignment
The late penalties will be applied in Canvas separately for the code and
, COMP3308/3608 AI, week 13b, 2022 2
Last week of the semester – we are approaching the finish line!
from Kenya wining the men’s marathon at the Rio Olympic Games in 2016
• How do you feel?
• happy, excited, tired,
exhausted, relieved …
• Do you think you have learned useful things in this course?
Images from: https://www.iaaf.org/news/report/rio-2016-men-marathon, Getty images
, COMP3308/3608 AI, week 13b, 2022 3
• Complete the Unit of Study Survey https://student-surveys.sydney.edu.au/students/
• Talk about the exam
, COMP3308/3608 AI, week 13b, 2022 4
Unit of study surveys – why you should do them?
• We really need you do complete the survey. Please do it now! https://student-surveys.sydney.edu.au/students/
• These surveys are required for all courses, not only for AI
• Please take them seriously and complete them properly – so many
important decisions are based on them, e.g.
• Courses with low student satisfaction are investigated by the Head of
School and Associate Dean Education
• Courses with high student satisfaction are commended
• There is also a requirement for a minimum response rate, so please complete the survey!
, COMP3308/3608 AI, week 13b, 2022 5
Why you should do them? (2)
We need to know what went well and what didn’t
We put a lot of effort in this course – we need your feedback We moved the assignments to Grok
We make changes every year based on the student surveys
If you had any difficulties or concerns, please be specific and explain them
You don’t want your comments to be dismissed, e.g. “It is natural students to complain when the material is difficult/not interesting/etc.”
You don’t want someone who doesn’t teach well to say: “I don’t know what the student meant; I did a good job…”
I can assure you that I read every single comment and make changes based on the surveys
, COMP3308/3608 AI, week 13b, 2022 6
Why you should do them? (3)
There are prizes!☺
Each completed survey will give you an entry into the Semester 1
USS prize draw:
• 1st prize – Apple iPad Air 64GB
• 2nd prize –
• 3rd prize – 2 x $200 JB Hi
• 4th prize – 6 x $100 JB Hi
• The draw will be on 18 July 2022 • Last semester’s winners:
https://student-surveys.sydney.edu.au/students/complete/prizes.cfm
, COMP3308/3608 AI, week 13b, 2022 7
Unit of study surveys (4)
The surveys are anonymous
We receive the results after all marks are finalised and don’t know which student gave
After the marks are finalised, you will receive a summary of the numeric ratings provided by all students in the course, together with my reply to your ratings and comments
Q6 is about feedback on your learning
Was your learning supported well, were your questions answered well on Ed and during
the tutorials; feedback on the assessments
F133 is about the tutorials. Write the name of your tutor (in the text comment
which rating or comment
section at the end):
• in-person: Tue 4pm and 5pm: Danielle (Carslaw)
• in-person: Tue 4pm: Cameron (Business ABS)
• in-person: Wed 10am and 11am: Vincent (Merewether and Business ABS) • in-person: Wed 1pm and 2pm: Shelley (Business ABS)
• in-person: Wed 4pm and 5pm: Nick (Business ABS)
• online: Tuesday 4pm: Christopher
• online: Tuesday 4pm and 5pm, Wed 3pm and 4pm: Stephen
• online: Wed 10am, 11am, 1pm: OMP3608: Cristopher (in-person, Wed 3pm, Business ABS)
Please complete the survey: https://student-surveys.sydney.edu.au/students/
, COMP3308/3608 AI, week 13b, 2022 8
TA: Christopher and recordings: will pass all your comments to your tutors!
• Online, on Canvas, set as a Quiz
• Proctored: Record+ (Type B exam)
• Duration: 2 hours + 10 min reading time = 130 min
• In addition: 15 min upload time
• Preparing for online exams: https://canvas.sydney.edu.au/courses/23380
• Help and FAQs about Record+ exams:
https://canvas.sydney.edu.au/courses/23380/pages/record+-help-and-faqs
• Record+ exam: you will be recorded via webcam and screen-sharing, and monitored by an AI software
• After the exam, a human proctor will review both the recording and any flags that were made by the AI software about suspicious activity. An incident report will be sent to the unit coordinator. Educational integrity decisions are always made by the university and not ProctorU.
, COMP3308/3608 AI, week 13b, 2022 9
Permitted Materials
Permitted materials during the exam (the proctoring session):
1 page of your own notes – double-sided A4 size, handwritten or typed
(paper-based, not electronic)
Calculator – handheld, non-programmable Blank scratch paper, multiple sheets
No other materials are allowed – no lecture slides, tutorial notes, books or other materials
No other devices are permitted
No internet browsing is allowed
You can’t consult other people during the exam
, COMP3308/3608 AI, week 13b, 2022 10
During Upload Time
For one of the questions (the last question), and only for this question, you can either:
(1) type your answers and working in the Quiz (there is box) or
(2) write them on paper, take a photo with your phone and upload the photo
If you choose the second option, the upload must be done during the upload time which is after the proctoring session has finished. The photo should be uploaded under “Assignments” in the Canvas exam site.
You are not allowed to use your phone during the proctoring session
, COMP3308/3608 AI, week 13b, 2022 11
Academic Honesty
• All suspicious behavior during the proctoring session will be reviewed
• All file uploads will be compared for plagiarism
• Please do not cheat or copy!
• The consequences and penalties are very severe
• If you copy, you will get caught. If you make your work available to another student to copy, this is also academic dishonesty and you will be investigated and penalized.
• The stress of going through the investigation is immense
• The investigation takes many months and your mark will not be finalised until it is completed, your graduation may be delayed, you may have problems enrolling in other courses.
• It is not worth it. You will regret it all your life.
, COMP3308/3608 AI, week 13b, 2022 12
Exam Paper is Confidential
• The exam paper is confidential
• You must not discuss the exam questions with other people, post or
distribute the exam questions in any way, during or after the exam
, COMP3308/3608 AI, week 13b, 2022 13
Canvas Exam Site
The Canvas site for the exam is different that the Canvas site we
use during the semester for teaching – 2 exam sites:
Final exam for: COMP3308 Final exam for: COMP3608
The Exams Office will give you access to your exam site
, COMP3308/3608 AI, week 13b, 2022 14
More Exam Advice
Check and double-check the day and time of your exam
Make sure you are ready for proctored Record+ exam
You need to have a ProctorU account, install ProctorU software and take a
practice test
• https://canvas.sydney.edu.au/courses/23380
• https://canvas.sydney.edu.au/courses/23380/pages/record+-help-and-faqs
Be calm and well organized
Allocate time according to marks for each question
You should find the exam easy if you:
• Have attended lectures and followed up
• Have attended tutorials, have done all tutorial exercises and followed up
• Have read the recommended reading from the textbooks
• Have done well in the assignments
, COMP3308/3608 AI, week 13b, 2022 15
Marks and Examinable Content
Marks total: 60 (=60% of your final mark)
A minimum of 24 marks (40%) on the exam is required to pass the course
All material is examinable except:
Week 1 – introduction
Week 13a (applications of AI – recommender systems) Historical context, Matlab and Weka
, COMP3308/3608 AI, week 13b, 2022 16
Types of Questions
The exam contains 3 types of questions:
Useful documents:
• Sample exam questions
• Search: Weeks_2-3_Practice.pdf
• Bayesian networks: BN_practice_questions.pdf
Type 1: questions requiring short answers – test your understanding and ability to relate concepts; be clear and concise
Type 2: calculation/problem solving
Type 3: multiple-choice questions (small number)
The exam questions are similar to the tutorial questions
, COMP3308/3608 AI, week 13b, 2022 17
Read the Questions Carefully
• Do what the question asks for:
• Some questions ask you to show your calculations/working or to
give an explanation – most of the questions fall into this category
• However, some other questions just ask you to show the final answer
, COMP3308/3608 AI, week 13b, 2022 18
Type 1: Short-Answer Questions
a) Gain ratio is a modification of Gain used in decision trees. What is its advantage?
It penalizes highly-branching attributes by taking into account the number and the size of branches.
b) Why do we need to normalize the attribute values in the k- nearest-neighbor algorithm?
As different attributes are measured on different scales, without normalization the effect of the attributes with smaller scale of measurement will be less significant than those with larger.
c) What is the main limitation of the perceptrons?
Can separate only linearly separable data.
, COMP3308/3608 AI, week 13b, 2022 19
Short-Answer Questions (2)
d) The 1R algorithm generates a set of rules. What do they test?
The values of a single attribute.
if outlook=sunny then play=no
elseif outlook=overcast then play=yes elseif outlook=rainy then play=yes
e) The problem of finding a decision boundary in SVM can be formulated as an optimisation problem using Lagrange multipliers. What are we maximizing? The margin of the hyperplane.
f) In linear SVM, we use dot products both during training and during classification of a new example. What vectors are these products of?
During training: Pairs of training examples.
During classification of a new example:
The new example and
the support vectors.
, COMP3308/3608 AI, week 13b, 2022 20
sunny overcast
rainy no yes yes
Short-Answer Questions – Backpropagation
• Your task is to develop a computer program to rate chess board positions
• You got an expert chess player to rate 100 different chessboard positions and then use this data to train a backpropagation neural network, using board features as the ones shown in the figure below:
, COMP3308/3608 AI, week 13b, 2022 21
Short-Answer Questions – Backpropagation (2)
Select the correct answer (“Yes” or “No”) and briefly explain it.
• Select “Yes” for all issues that could, in principle, limit your ability to develop the best possible chess program using this method. Select “No” for all issues that could not.
1. The backpropagation network may be susceptible to overfitting of the training data, since you tested its performance on the training data instead of using cross validation.
Testing the performance on the training data is an over-optimistic measure and it does not guard against overfitting. Cross-validation is a more reliable procedure.
2. The backpropagation neural network can only distinguish between board positions that are completely good or completely bad.
The backpropagation neural network can be used for both classification and regression tasks. In regression tasks the output is numeric and can represents different levels of good and bad.
, COMP3308/3608 AI, week 13b, 2022 22
Short-Answer Questions – Backpropagation
3. The backpropagation network will converge to the global minimum. No
The backprogarion algorithm implements gradient descent and is not guaranteed to converge to the global minimum – it finds the closest local minimum.
4. You should have used higher learning rate and momentum to guarantee convergence to the global minimum.
The use of momentum reduces the oscillations when using a higher learning rate but it doesn’t guarantee convergence to the global minimum.
5. The topology of your neural net might not be adequate to capture the expertise of the human expert.
Too few neurons – underfitting; too many – overfitting.
, COMP3308/3608 AI, week 13b, 2022 23
Type 2: Problem Solving Question – Nearest Neighbor
• In the figure below, the circles are training examples and the squares are test examples, i.e. we are using the circles to predict the squares
• Two algorithms are used: 1-Nearest Neighbor and 3-Nearest Neighbor
, COMP3308/3608 AI, week 13b, 2022 24
Problem Solving Question – Nearest Neighbor
We know that:
What will be the class of the following examples? Write +, – or U for cannot be determined.
Square 6 using 1-Nearest Neighbour: Square 6 using 3-Nearest Neighbour: Square 3 using 1-Nearest Neighbour: Square 5 using 1-Nearest Neighbour:
, COMP3308/3608 AI, week 13b, 2022 25
We know that:
What will be the class of the following examples? Write +, – or U for cannot be determined.
Circle L: –
Circle I: +
Circle H: –
Circle E: +
Circle K: U
Circle C: +
Square 6 using 1-Nearest Neighbour: U Square 6 using 3-Nearest Neighbour: – Square 3 using 1-Nearest Neighbour? + Square 5 using 1-Nearest Neighbour? U
, COMP3308/3608 AI, week 13b, 2022 26
Problem Solving Question – Search
In the tree below the step costs are shown along the edges and the h values are shown next to each node. The goal nodes are double-circled: F and D.
Write down the order in which nodes are expanded using:
e) Greedy search
In case of ties, expand the nodes in alphabetical order.
, COMP3308/3608 AI, week 13b, 2022 27
Search – Review
• BFS: Expands the shallowest unexpanded node
• DFS: Expands the deepest unexpanded node
• UCS: Expands the node with the smallest path cost g(n) (from the root!)
• IDS: DFS at levels l = 0, 1, 2, etc.; expands the deepest unexpanded node within level l
• Greedy: Expands the node with the smallest heuristic value h(n)
• A*: Expands the node with the smallest f(n)=g(n)+h(n)
, COMP3308/3608 AI, week 13b, 2022 28
b) DFS: ABEF •
c) UCS: ABEF
d) IDS : AABCD
e) Greedy search: AD
f) A*: ABEF
COMP3308/3608 AI, week 13b, 2022 29
• Write down the order in which nodes are expanded using:
Read the question – no explanation is needed, just list the nodes
Short Answers – Search
• A* uses admissible heuristics. What happens if we use a non- admissible one? Is it still useful to use A* with a non-admissible heuristic?
Not optimal anymore. But it could still find a reasonably good solution in acceptable time, depending on how good the heuristic is.
• What is the advantage of choosing a dominant heuristic in A* search?
Expands a fewer number of nodes. As a result, the optimal solution will be found quicker.
• What is the main advantage of hill climbing search over A* search? Space complexity – keeps only the current node in memory
, COMP3308/3608 AI, week 13b, 2022 30
Where to Apply AI and ML?
• IBM’s big 5 ML application areas:
1. Education – the classroom of the future will “learn” the student
2. Shopping – personal shopping, predicting what customers will like
3. Medicine – personalised disease treatment based on your DNA
• matching patients with other patients (what worked for them), predicting the right diagnosis and treatment
4. Security – fraud detection and protection (bank accounts, passwords on social media, etc.)
5. Smart cities – understanding and responding to citizen’s needs (transport, emergencies, housing, electricity consumption; data from crowdsourcing, mobile applications and sensors)
• Application areas aligned with UN development goals – e.g. quality education, good health, 0 hunger and homelessness, clean water, etc. http://www.un.org/sustainabledevelopment/sustainable-development-goals/
• Workshop on Data Science for Social Good at the European Conference on ML & DM: https://sites.google.com/view/ecmlpkddsogood2022/
, COMP3308/3608 AI, week 13b, 2022 31
The Last Slide
• We hope that you found this course useful
• Use what you have learned in this course!
• Apply AI and ML – there are so many opportunities!
• There are already some great AI and ML success stories, but there will be even greater ones in the future and you can be part of them! You can contribute to them or lead them!
All the best!
Christopher, Nick, James, Danielle, Shelley, Stephen, Cameron, Vincent and Irena
, COMP3308/3608 AI, week 13b, 2022 32
程序代写 CS代考 加微信: powcoder QQ: 1823890830 Email: powcoder@163.com