CS代考 VERSION 00000001 COMPSCI 367

VERSION 00000001 COMPSCI 367
THE UNIVERSITY OF AUCKLAND
SEMESTER TWO, 2015 Campus: CITY
COMPUTER SCIENCE
Artificial Intelligence (Time Allowed: TWO hours)
You must answer all questions in this exam.
No calculators are permitted
Answer Section A (Multiple choice questions) on the Teleform answer sheet provided. Answer Sections B and C in the space provided in this booklet.
There is space at the back for answers that overflow the allotted space.
Surname Forename(s) Student ID Login (UPI)
1 – 18 19 – 25 26 – 33
Multiple Choice Section B Section C
Page 1 of 23

VERSION 00000001
COMPSCI 367
SECTION A MULTIPLE CHOICE QUESTIONS
Each question in this section is worth 2 marks. There is only one correct answer for each question. For each question, choose the best answer according to the information presented in lectures. Select your preferred answer on the Teleform answer sheet provided by shading in the appropriate box.
Question 1
[2 marks] A descriptive theory specifies how something occurs in nature (e.g., how humans behave), whereas a prescriptive theory specifies how one should do something. Which of these statements about theories of decision making is true?
(a) Neither decision theory or heuristic choice theory are prescriptive.
(b) Decision theory is prescriptive and heuristic choice theory is descriptive. (c) Both decision theory and heuristic choice theory are descriptive.
(d) Decision theory is descriptive and heuristic choice theory is prescriptive. (e) None of the above.
Question 2
[2 marks] Suppose one is given a set of alternatives A1, …, Am , a set of attribute values Vj1, …, Vjn for each Aj , and a set of N weights W1, …, Wn. Which of these expressions describes a linear utility function Uj of the type used in decision theory?
(a)Uj= !!!W!×V!” (b)Uj= !!!W!×V!”
(c)Uj= !!!W!+V!” (d)Uj= !!!W!+V!” (e)Uj= !!!W!×V!”
Question 3
[2 marks] Heuristic choice theory is a well-known framework for decision making. Which of these statements is NOT an assumption of the theory?
(a) When making choices, people ‘satisfice’ in that they select alternatives which they find good enough.
(b) Humans draw on ‘rules of thumb’ that generally produce acceptable results with limited effort.
(c) Domain-specific heuristics are linked to chunks, which are relational patterns with associated responses.
(d) When making choices, people consider all attributes of each alternative and make the best selection.
(e) None of the above.
Page 2 of 23

VERSION 00000001 COMPSCI 367
Question 4
[2 marks] The stimulus-response paradigm is widely adopted in control systems and is central to behaviorist theories in psychology. Which of these aspects of human activity do stimulus- response systems support?
(a) Taking goals into account when deciding on actions.
(b) Using memories about recent decisions and events.
(c) Matching long-term structures to select appropriate actions. (d) Utilizing knowledge about the order of activities.
(e) Organizing complex activities into subactivities.
Question 5
[2 marks] Hierachical task networks (HTNs) offer a way to specify knowledge about plans in a given domain, which they encode as a set of methods. Which of these is NOT an element of an HTN method?
(a) A set of subtasks, with arguments, that implement the method. (b) A set of conditions under which the method may be applied. (c) A task predicate and its associated arguments.
(d) An optional set of effects that the method produces.
(e) A set of other methods that may occur after its application.
Question 6
[2 marks] Production systems are used to model human cognition and construct knowledge- based systems, but they also incorporate ideas from stimulus-response theories. Which of these assumptions is NOT shared by the stimulus-response and production-system paradigms?
(a) Knowledge is stored as a set of conditional patterns with associated responses.
(b) Behavior relies on matching patterns to determine appropriate actions.
(c) Sequential information processing operates in discrete cycles.
(d) Patterns can match against either external perceptions or internal mental structures. (e) There is no explicit hierarchical organization of knowledge.
Question 7
[2 marks] Production systems are often used to model high-level cognition in humans. Which of these is NOT a reason they are used for that purpose?
(a) They offer an easy way to model perceptual-motor activities. (b) They balance stimulus-driven and goal-driven behavior.
(c) They combine parallel retrieval with sequential decisions.
(d) They provide a good level of analysis for many symbolic tasks. (e) Their modular representation supports structural learning.
Page 3 of 23

VERSION 00000001 COMPSCI 367
Question 8
[2 marks] The recognize-act cycle is central to the operation of any production system. Which of these is NOT a step in this cycle?
(a) Finding all ways that each rule’s condition side matches against the contents of working memory.
(b) Using conflict resolution to select one or more matched rule instances to execute.
(c) Invoking the next production rule to execute in a procedural sequence.
(d) Instantiating and carrying out the actions of selected matches.
(e) Checking to see whether any rules match against working memory and halting if not.
Question 9
[2 marks] A qualitative causal model specifies a set of causal influences between pairs of quantitative variables. Based on the causal model depicted above, which effects will happen if we decrease B?
(a) E will increase and F will increase. (b) E will decrease and F will increase. (c) C will increase and F will decrease. (d) A will decrease and D will increase. (e) None of the above.
Question 10
[2 marks] One can use a qualitative causal model of a dynamic system to generate an envisionment that describes its behavior over time. Which of these elements is NOT part of such an envisionment?
(a) A set of possible qualitative states.
(b) Numeric values for variables at each time point.
(c) An initial qualitative state.
(d) Transitions that lead from states to their successors. (e) None of the above.
Page 4 of 23

VERSION 00000001 COMPSCI 367
Question 11
[2 marks] There are many similarities between deductive and abductive inference. Which of these features is NOT true for both of these forms of reasoning?
(a) They make plausible default assumptions in order to complete explanations. (b) They operate over knowledge stated as rules or similar structures.
(c) They carry out search to find proof trees.
(d) They are provided with known or observed ground literals.
(e) They generate proof trees composed of linked rule instances.
Question 12
[2 marks] The tasks of plan understanding and activity recognition share some characteristics but not others. Which of these features do they have in common?
(a) They utilize abductive inference rather than classification.
(b) They attempt to construct an explanation in terms of the agent’s plans or goals. (c) They accept as input a sequence of states traversed or actions taken by an agent. (d) They introduce plausible default assumptions during processing.
(e) None of the above.
Question 13
[2 marks] Case-based reasoning and analogical reasoning have many features in common. Which of these processing stages is NOT shared by these two approaches?
(a) Chaining rule instances to construct an explanation. (b) Adapting a selected case to perform some task.
(c) Retrieving relevant cases from memory.
(d) Selecting a retrieved case to utilize.
(e) None of the above.
Question 14
[2 marks] Which of these features does NOT distinguish analogical reasoning and abductive inference from deductive reasoning?
(a) Analogy and abduction illustrate that symbolic methods need not be fragile. (b) Analogy and abduction operate over symbolic, relational structures.
(c) Analogy and abduction use partial matching rather than all-or-none matching. (d) Analogy and abduction use plausible inference rather than deductive proofs. (e) None of the above.
Page 5 of 23

VERSION 00000001 COMPSCI 367
Question 15
[2 marks] Which of these statements is NOT consistent with Clark’s (1996) claims about common ground in dialogue?
(a) Common ground expands incrementally as the conversation proceeds.
(b) Common ground includes assumptions shared by participants at the start of a
conversation.
(c) Common ground includes beliefs about others’ beliefs and goals.
(d) Much of common ground is implicit and must be inferred over the course of a
(e) None of the above.
Question 16
[2 marks] Which of these cognitive abilities is NOT central to human-level dialogue processing?
(a) Representing other agents’ beliefs and goals.
(b) Understanding and generating sentences.
(c) Drawing domain-level and dialogue-level inferences. (d) Using utterances to alter others’ mental states.
(e) None of the above.
Question 17
[2 marks] Appraisal theory is a major account of emotions that makes close contact with ideas from artificial intelligence. Which of these statements is NOT consistent with appraisal accounts of emotion?
(a) Many emotions involve inferences about others’ mental states.
(b) Emotional structures are abstract and domain independent.
(c) Emotions are primarily about one’s visceral feelings.
(d) Emotions involve relations among goals, expectations, and beliefs. (e) None of the above.
Question 18
[2 marks] Personality refers to stable regularities in an agent’s behavior. Which of the following is NOT consistent with the personality metacognition hypothesis?
(a) Personality-related structures inspect traces of basic cognition and alter its course.
(b) Personality involves long-term structures that specify the conditions for generating
(c) Personality-related structures modulate emotional responses.
(d) Personality can be described as a point in a trait space of N fixed dimensions.
(e) None of the above.
Page 6 of 23

VERSION 00000001 COMPSCI 367
THIS PAGE HAS BEEN INTENTIONALLY LEFT BLANK.
Page 7 of 23

VERSION 00000001 COMPSCI 367
THIS PAGE HAS BEEN INTENTIONALLY LEFT BLANK.
QUESTION/ANSWER BOOKLET FOLLOWS Page 8 of 23

VERSION 00000001 COMPSCI 367 QUESTION/ANSWER BOOKLET ID…………………..
SECTION B: ‘s material
Question 19
What is the main difference between a local search algorithm and a global search algorithm? [4 marks]
Question 20
What is the main purpose of alpha-beta pruning in the minimax algorithm? [4 marks]
Question 21
What is the difference between overfitting, underfitting, and early convergence in supervised machine learning? [4 marks]
Page 9 of 23

VERSION 00000001 COMPSCI 367 QUESTION/ANSWER BOOKLET ID…………………..
Question 22
What benefit do you get from running genetic algorithms or neural networks multiple times on the same training set as opposed to running something like a decision-tree algorithm or the candidate elimination algorithm multiple times on the same training set? [4 marks]
Question 23
Why should you not report only the accuracy of your classifier on the training set? [4 marks]
Page 10 of 23

VERSION 00000001 COMPSCI 367 QUESTION/ANSWER BOOKLET ID…………………..
Question 24
What is the difference between the hypothesis space searched by a neural network algorithm versus the hypothesis space searched by a decision trees algorithm? [4 marks]
Question 25
What are the main properties of swarm intelligence that gives it its power? [4 marks]
Page 11 of 23

VERSION 00000001 COMPSCI 367 QUESTION/ANSWER BOOKLET ID…………………..
PART C: ’s material
Question 26
Write English sentences for the following first-order expressions. [6 marks] ∀x∃y likes(x,y) ∧ likes(y,icecream)
∃y∀x likes(x,y) ∧ likes(y,icecream)
Page 12 of 23

VERSION 00000001 COMPSCI 367 QUESTION/ANSWER BOOKLET ID…………………..
Question 27
Write Prolog code to define grandmother. This is either your mother’s mother or your father’s mother. Do this using only: grandmotherOf(X,Y) (which you are defining), motherOf(X,Y), and fatherOf(X,Y). [5 marks]
Page 13 of 23

VERSION 00000001 COMPSCI 367 QUESTION/ANSWER BOOKLET ID…………………..
Question 28
What is the main advantage of using weighted A* over using A* and what is the main disadvantage? [5 marks]
Page 14 of 23

VERSION 00000001 QUESTION/ANSWER BOOKLET
Question 29
Describe perimeter search. Specifically, describe:
A. How is the perimeter grown?
B. What does the heuristic try to “measure”? C. When does perimeter search terminate? D. Is the solution guaranteed to be optimal?
COMPSCI 367 ID…………………..
Page 15 of 23

VERSION 00000001 COMPSCI 367 QUESTION/ANSWER BOOKLET ID…………………..
Question 30
What is the relationship between planning and search? Think about the search assignment (8 puzzle) and the planning assignment (RushHour). How were states represented in both and how were actions represented in both? [5 marks]
Page 16 of 23

VERSION 00000001 COMPSCI 367 QUESTION/ANSWER BOOKLET ID…………………..
Question 31
Assuming the cost of regressing goals through operators is approximately the same as the cost of progressing states through operators, what feature of the search space would determine whether it would be better to do forward state space search or backward state space search? Why? [4 marks]
Page 17 of 23

VERSION 00000001 COMPSCI 367 QUESTION/ANSWER BOOKLET ID…………………..
Question 32
Comparing plan space planning with state space planning, what are plan space planners attempting to do better than state space planners? What current shortcoming of plan space planning has led state space planners to be today’s dominant planners? Why?
Page 18 of 23

VERSION 00000001 COMPSCI 367 QUESTION/ANSWER BOOKLET ID…………………..
Question 33
In our planning assignment, how were meta-level predicates different than derived-level predicates? Specifically, what appeared in the body of meta-level predicates that did not appear in the body of derived predicates and vice versa. Why is this? [3 marks]
Page 19 of 23

VERSION 00000001 QUESTION/ANSWER BOOKLET
COMPSCI 367 ID…………………..
– Overflow Sheet 1 –
Write the question number and letter next to your answer. You must ALSO indicate in the allotted space that you have used the overflow sheet.
Page 20 of 23

VERSION 00000001 QUESTION/ANSWER BOOKLET
COMPSCI 367 ID…………………..
– Overflow Sheet 2 –
Write the question number and letter next to your answer. You must ALSO indicate in the allotted space that you have used the overflow sheet.
Page 21 of 23

VERSION 00000001 COMPSCI 367 QUESTION/ANSWER BOOKLET ID…………………..
Rough Working – This page will not be marked
Page 22 of 23

VERSION 00000001 COMPSCI 367 QUESTION/ANSWER BOOKLET ID ………………….
Rough Working – This page will not be marked
________________________________________
Page 23 of 23