Download Tweak Operators - Artificial Intelligence - Exam and more Exams Artificial Intelligence in PDF only on Docsity!
CS 540-1 -- Introduction to Artificial Intelligence
Final Exam - December 16, 1992
(Rm 6203, Soc Sci Building)
CLOSED BOOK
120 minutes (5:05-7:05 pm)
Write your answers on these pages and show your work. If you feel that a question is not fully specified, state any assumptions you need to make in order to solve the problem. You may use the backs of these sheets for scratch work.
Write your name on this and all other pages of this exam. Make sure your exam contains eight problems on ten pages.
Name ________________________________________________
Student ID ________________________________________________
Problem Score Max Score
1 _____ 20
2 _____ 15
3 _____ 35
4 _____ 15
5 _____ 15
6 _____ 20
7 _____ 15
8 _____ 15
Total _____ 150
PROBLEM 1 - Non-Linear Planning (20 points)
Consider the following TWEAK operators:
Operator 1 Operator 3 preconditions: A ∧ ¬ C preconditions: ¬ B postconditions: ¬ A ∧ C postconditions: Y ∧ ¬ X ∧ ¬ Z Operator 2 Operator 4 preconditions: ¬ A ∧ C preconditions: ¬ C postconditions: X ∧ A postconditions: ¬ Y ∧ Z
Assume the task is to construct a plan that correctly converts the initial state A ∧ ¬ B ∧ ¬ C into the final state X ∧ Y ∧ ¬ Z.
Show a sequence of partial plans that TWEAK could produce during its search for an acceptable non-linear plan. When choosing goals from the list of unachieved goals, always choose the one earliest in the alphabet. Briefly explain the reason for each of TWEAK’s alterations of the current partial plan.
PROBLEM 3 - Neural Networks and Inductive Learning (35 points)
Part A. Show that performing gradient descent for the perceptron leads to the delta rule for weight adjustment. You may assume that the activation function for the output unit is f(x)=x , i.e. the value of the output unit is simply the weighted sum of its inputs. Also, assume the error measure is one-half the square of the difference between the teacher’s and the network’s values.
Part B. Describe the spaces searched by backpropagation and ID3.
Part C. Which search-control strategy does backpropagation use (e.g., depth-first search, beam search, best-first search, etc.)? What about ID3? Explain your answers.
PROBLEM 4 - Waltz Labeling (15 points)
cover this with the figure of label types from the book!!!!!!!
Label the line segments in the figure below using the vertex labels provided above. Assume the outermost segments are figure/background boundaries.
PROBLEM 6 - Miscellaneous Questions (20 points)
Answer the following true (T) or false (F). Provide brief justifications of your answers.
hhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhh
If hill climbing finds a solution path, S-P, beam search will also produce S-P as its answer.
Ans: _____ Just:
hhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhh
Best-first search is guaranteed to find an optimal solution.
Ans: _____ Just:
hhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhh
F(F(?x, k, ?y), h, ?y) and F(?z, ?x, G(?x)) unify.
Ans: _____ Just:
hhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhh
∀ x F(x) is logically equivalent to ¬∀ x ¬ F(x).
Ans: _____ Just:
hhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhh
The Waltz labeling procedure produces exactly one interpretation for each drawing.
Ans: _____ Just:
hhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhh
The use of scripts can aid the task of natural language understanding.
Ans: _____ Just:
hhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhh
Viewer-center representations require less storage than object-oriented ones.
Ans: _____ Just:
hhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhh
PROBLEM 7 - Important AI Concepts (15 points)
In the space below, provide brief and succinct summaries of why the following are important. You need to only summarize five (5) of the seven (7) concepts. Be sure you clearly indicate which five you are answering.
hhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhh
resolution theorem proving
hhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhh
A*
hhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhh
iterative deepening
hhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhh
ISA hierarchies
hhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhh
frame axioms
hhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhh
test sets (in machine learning)
hhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhh
hidden units in a neural network
hhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhh
Part B. Write a Prolog program for the predicate double_list(List, Doubled_list). This predicate is true when Doubled_list contains an item-by-item duplicated version of List , as illustrated below:
hhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhh
?- double_list([a,b,c], [a,a,b,b,c,c]). yes
?- double_list([a,b,c], X). X = [a,a,b,b,c,c]
?- double_list(X, [a, a, b, b]). X = [a,b]
hhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhh
The end - have a good vacation!
