Submission deadline: 23:59, March 31st, 2023

Percentage in final mark: 50%

Learning outcomes assessed: A, B

Individual/Group: Individual

Length: The assessment has a total of 8 questions which gives 100 marks. The submitted file must be in pdf format.

Late policy: 5% of the total marks available for the assessment shall be deducted from the assessment mark for each working day after the submission date, up to a maximum of five  working days

Risks:

•  Please read the coursework instructions and requirements carefully. Not following these instructions and requirements may result in loss of marks.

•  The formal procedure for submitting coursework at XJTLU is strictly followed. Submission    link on Learning Mall will be provided in due course. The submission timestamp on Learning Mall will be used to check late submission.

__________________________________________________________________

Question 1- Agents (10 marks):

Consider the task of assembling a bookcase from its components.

(a) Use PEAS to specify the parameters for the task environment.

(i) Performance measure:

(ii) Environment:

(iii) Actuators:

(iv) Sensors: (4 marks)

(b) Specify three properties of the environment. Explain these properties in the context

of the task (no marks will be given for an absent or incorrect explanation). (3 marks)

(c) Determine which of the following agent types would you employ: Simple reflex, Model    based, Goal based or Utility based. Explain why in the context of the task, and list your assumptions (no marks will be given for an absent or incorrect                 explanation). (3 marks)

Question 3- DFS/BFS (10 marks):

Use the following (fully expanded) search tree to indicate the order in which nodes are          expanded for different types of search. Assume that A is the start node and G (double boxed) the only goal node. How many nodes are there in the resultant search tree?

(a) Depth-first search

List the nodes according to their order of expansion.

List the nodes in the final search tree (without the nodes deleted by the algorithm). (3 marks)

(b) Best-first greedy search

List the nodes according to their order of expansion. For each expansion, list OPEM (with the nodes in the correct order) and CLOSED.

List the nodes in the final search tree. (7 marks)

Question 4- Algorithm A/A* (15 marks):

The following (fully expanded) search tree indicates the cost of reaching each node  (g). the     estimated cost of reaching a goal node from each node  (h), and their sum (f). For example, the cost of reaching node C is 13, and the estimated cost of reaching the goal node from node C is 17. A is the start node, and nodes with 0 cost estimates  (h)are goal nodes.

(a) Draw the search tree that would be generated by Algorithm A, using roman numerals to indicate the order in which nodes are expanded. Upon completion, mark the goal node. After each expansion, provide the following information in the table below:

•    the node that was expanded, and

•    the nodes in the horizon (OPEN) in order of merit, i.e., in the order in which they will

be expanded, together with their f value. If several nodes have the same merit, order them from left to right. (13 marks)

(b) Did algorithm A find the optimal answer, and is the h function admissible? Why or why not? Illustrate your answer with an example from the provided search tree    (no marks will be given for an absent or incorrect explanation). (2 marks)

Question 5- Simulated Annealing/Genetics Algorithms (10 marks):

(a) In Simulated Annealing, if T2 > T1, is the probability of adopting a new worse state higher in T2 or in T1? Why? (no marks will be given for absent or incorrect explanations) (4 marks)

(b) Does the Simulated Annealing algorithm given in class always terminate? Why or why not? (no marks will be given for absent or incorrect explanations) (3 marks)

(c) A genetic algorithm is used to evolve a binary string of length n to one where the sum (from left to right) of the last four genes is equal to 1. The initial population is a randomly generated  set of binary strings of length n, such as those shown here:

00110001

01011101

11101111

Give a suitable fitness function for this problem. (3 marks)

(b)       What is the probability that a hamburger eater will have Creutzfeldt-Jakob disease?

Show the formula before you plug-in numbers. If you don't have a calculator, you can still estimate the answer. (4 marks)

(c)       Should you stop eating hamburgers? Why or why not? (no marks will be given for an absent or incorrect explanation). (2 marks)

2)        Consider a probabilistic model (Bayesian Network) for determining a patient’s risk of

heart disease. The model consists of the following 5 random variables:

H: Heart Disease = {present, absent}

B: Blood Pressure = {high, low}

E: Exercise = {frequent, infrequent}

G: Gender = {male, female}

S: Smoking = {smoker, non-smoker}

The model admits the following factorisation of the joint distribution for the 5 variables: P(H, B,

E, G, S) = P(H|B, E, G, S)P(B|E, G)P(E)P(G)P(S)

a)        If we don’t know anything about the patient, is Smoking (S) independent of Exercise (E) and why? (1.5 marks)

b)        Suppose you are told that the patient has high Blood Pressure (B). Is Smoking (S) independent of Exercise (E)? (1.5 marks)

c)        And if you are told that the patient has Heart Disease (H) are S and E independent? (1.5 marks)

d)       You know from your medical training that frequent exercise lowers the risk of heart  disease while smoking increases it (irrespective of gender). If you are told that a       particular patient with the disease was NOT a smoker, what can you infer about their level of exercise? (1.5 marks)

Question 8: Decision Tree (17 marks):

You regularly take the Suzhou to Taicang shuttle bus. You want to build a Decision Tree to predict whether you will arrive on time, and you have gathered the following set of six training instances: