CSC384代做、c++程序设计调试、Java，Python语言代写

日期：2020-04-11 11:05

Take Home Exam , University of Toronto, CSC384 - Intro to AI, Winter 2020 1

Computer Science 384 April 3, 2020

Take Home Exam: Bayes Nets and Knowledge Representation

Due: April 17, 2020 by 10:00 PM(EDT)

Silent Policy: A silent policy will take effect 24 hours before this assignment is due, i.e. no questions will

be answered, whether asked on the discussion board, via email or in person.

Policies:

1. The TAs and instructors will continue to hold office hours and host help sessions between April 3rd

and the due date. However, during these sessions, you may not discuss problems on the take home

exam. Instead, you can discuss practice problems that have been posted to the website. Similarly,

on Piazza, you may not discuss problems on the take home exam. You can instead discuss practice

problems.

2. You must work alone on this take home exam. You may not discuss problems on the take home

exam with anyone (including other students).

3. You must write your answers clearly and legibly for full marks.

4. No submissions will be accepted past the due date without approval.

5. There will be no auto-fail policy associated with this exam.

Total Marks: This exam represents 20% of the course grade.

Handing in this Assignment

What to hand in electronically: Submit written answers in a file called answers.pdf as well as

acknowledgment form.pdf using MarkUs. Your login to MarkUs is your teach.cs username and password.

It is your responsibility to include all necessary files in your submission.

Clarification Page: Important corrections (hopefully few or none) and clarifications to the assignment

will be posted on the Exam Clarification page, linked from the CSC384 web page, also found at: http:

//www.teach.cs.toronto.edu/~csc384h/winter/tests.html. You are responsible for monitoring

the Exam Clarification page.

Questions: Questions about the exam should be asked on Piazza:

https://piazza.com/utoronto.ca/winter2020/csc384/home.

You may also reach out to the TAs or one of the instructors. Please place ”Exam” and ”CSC384” in the

subject line of your email.

Take Home Exam , University of Toronto, CSC384 - Intro to AI, Winter 2020 2

Q1. Probability (worth 15/100 marks)

1. (worth 2 marks) There is a type of skin cancer that affects 3 in every 100 people. A company has

invented a test that can diagnose this cancer using an image. The test isn’t perfect, tho; it will give a

false positive (i.e. it will detect cancer when there is none) 5% of the time and a false negative (i.e.

it will fail to detect a cancer that is present) 3% of the time.

If a test is positive, what is the probability the patient does not have cancer? If a test is negative,

what is the probability the patient does have cancer?

2. (worth 3 marks) Doctors are not happy with the false positive rate of the test. The company responds

by creating a new test that has a false positive rate of 6% and false negative rate of 4%. Although the

test seems worse than the original, the company explains the test results are conditionally independent

of one another given the condition of the user. They suggest using both tests in conjunction to

improve the false positive rate. Specifically, they suggest doctors diagnose cancer if and only if both

tests are positive. Does this logic make sense? Explain.

3. (worth 5 marks) We briefly discussed what it might mean to create an ’unbiased’ Bayesian Classifier.

Specifically, we said that if C is a classification, Y is a label representing ’ground truth’ and

A is some ’protected attribute’ (e.g. gender or race) we might enforce Separation of classifications,

making A independent of C given Y . Alternately we might enforce Sufficiency, making A is independent

of Y given C. But we can’t do both at the same time! Show that this is true, i.e. that

enforcing both Separation and Sufficiency implies A is independent of (Y, C).

4. (worth 5 marks) Given that X is independent of Y given Z and X is independent of W given (Y, Z).

Show that X is independent of (Y, W) given Z.

Take Home Exam , University of Toronto, CSC384 - Intro to AI, Winter 2020 3

Q2. Variable Elimination (worth 13/100 marks)

Birds frequently appear in the tree outside of your window in the morning and evening; these include

finches, cardinals and robins. Finches appear more frequently than robins, and robins appear more frequently

than cardinals (the ratio is 7:4:1). The finches will sing a song when they appear 7 out of every 10

times in the morning, but never in the evening. The cardinals rarely sing songs and only in the evenings

(in the evening, they sing 1 of every 10 times they appear). Robins sing once every five times they appear

regardless of the time of day. Every tenth cardinal and robin will stay in the tree longer than five minutes.

Every fourth finch will stay in the tree longer than five minutes.

1. (worth 2 marks) Draw a Bayesian network that captures the information in the story above correctly

and concisely. Make sure to annotate your network with conditional probability tables (CPTs).

2. (worth 1 mark) How many parameters will be required to specify the network you have drawn?

Explain.

3. (worth 5 marks) A bird lands in the tree in the morning. What is the probability that it will stay in

the tree longer than five minutes?

4. (worth 5 marks) What is the overall probability that any given bird in your tree will sing a song?

Q3. Markov Models (worth 12/100 marks)

In the mail room of a university office, a stream of blue and yellow mail carts pass a mail sorting robot. Blue

carts contain 7 domestic letters for every 3 international letters. Yellow carts contain 2 domestic letters for

every 3 international letters. Blue carts are followed by blue carts 70% of the time, but 30% of the time they

are followed by a yellow cart. Yellow carts are followed by yellow carts half of of the time, and the other

half of the time they are followed by a blue cart. The first cart, every morning, is yellow 9 of every 10 times.

The robot pulls a single letter from each cart as the cart passes it by.

1. (worth 2 mark) Draw a markov model to represent the joint distribution over cart colors and the

origins of letters selected by the robot. Include a conditional probability table (CPT) with each

variable in your model.

2. (worth 5 marks) What is the probability that the first three letters selected by the robot, in order,

will be: domestic, international, domestic?

3. (worth 5 marks) What is the probability that the fourth cart will be blue if the first three letters

selected, in order, are: international, international, domestic?

Take Home Exam , University of Toronto, CSC384 - Intro to AI, Winter 2020 4

Q4. D-Separation and Relevance (worth 10/100 marks)

Given the Bayesian Network structure above ....

1. (worth 1 mark) How many parameters are required to fully specify the network? Explain.

2. (worth 3 marks) List three pairs of variables X, Y where X is independent of Y .

3. (worth 3 marks) List three sets of variables X, Y, Z where X is independent of Z given Y .

4. (worth 1 mark) Assume we are to calculate P(B|E, F). Which variables are relevant?

5. (worth 2 marks) Assume we are to calculate P(B|G) using variable elimination. List the elimination

order you might suggest when using the min-fill heuristic to select variables, and give the size

of the factors that result from each elimination.

Take Home Exam , University of Toronto, CSC384 - Intro to AI, Winter 2020 5

Q5. First-order Structures and Models (worth 19/100 marks)

Consider a first-order language Lmetal consisting of constant symbols o1, o2, o3, a binary predicate symbol

heavier than, and unary predicate symbol expensive.

Let D = {Steel, Aluminium, T itanium}.

1. (worth 6 marks) How many Lmetal-structures with domain D exist? Justify your answer.

2. (worth 3 marks) Let Φ1 be the set consisting of the following sentences:

heavier than(o2, o1)

heavier than(o1, o3)

expensive(o1)

expensive(o2)

Consider a set B of Lmetal-structures for all structures M ∈ B:

hAluminium, Steeli ∈ heavier thanM

hSteel, T itaniumi ∈ heavier thanM

hT itanium, Aluminiumi ∈ heavier thanM

Aluminium ∈ expensiveM

Steel ∈ expensiveM

Are all structures in B models of Φ1? Explain why or why not.

3. (worth 4 marks) Suppose Φ2 is the set obtained by adding the following sentence to Φ1

?x?y?z(heavier than(x, y) ∧ heavier than(y, z)) → heavier than(x, z)

Are all structures in B models of Φ2? Explain why or why not.

4. (worth 6 marks) Consider the set Φ3 that contains only the following axiom

?x?y?z(heavier than(x, y) ∧ heavier than(y, z)) → heavier than(x, z)

How many Lmetal-structures with domain D are models of Φ3? Justify your answer.

Take Home Exam , University of Toronto, CSC384 - Intro to AI, Winter 2020 6

Q6. Proof by Resolution (worth 31/100 marks)

Consider the following knowledge base (note that p, p1, p2, p3 are constant symbols, part(x, y) means x

is a part of y and precedes(x, y) means x precedes y):

?c1?c2?c3

component(c1, p1) ∧ component(c2, p2) ∧ component(c3, p3)

∧ assemble before(c1, c2, p) ∧ assemble before(c2, c3, p)

(1)

?c1?c2(component(c1, p1) ∧ component(c2, p2) ∧ assemble before(c1, c2, p))→ ?c3 (component(c3, p3) ∧ assemble before(c1, c3, p) ∧ assemble before(c3, c2, p))(2)?c1?c2?a1?a2(component(c1, a1) ∧ component(c2, a2) ∧ assemble before(c1, c2, p))→ precedes(a1, a2)

(3)

?c1?c2?c3?a(assemble before(c1, c2, a) ∧ assemble before(c2, c3, a))

→ assemble before(c1, c3, a)

(4)

?c1?c2?a?a1(assemble before(c1, c2, a) ∧ component(c1, a1)) → part(a1, a)

(5)

1. (worth 11 marks) Convert the sentences to clausal form.

2. (worth 20 marks) Use resolution to answer the following queries.

You must use the notation developed in class (see slide no 39 in KRR-Part 2) for presenting your

answers.

(a) (worth 10 marks) What is a part of p? (finding one answer is sufficient)

Note: part(x, y) denotes x is a part of y.

(b) (worth 10 marks) What does precede p3? (finding one answer is sufficient)

Note: precedes(x, y) denotes x precedes y.

HAVE FUN and GOOD LUCK!