Data Structures and Algorithms – 2018

Assignment 2 – 25% of Continuous Assessment Mark

Deadline : 5pm Monday 12th February, via Canvas

Question 1 (22 marks)

It is sometimes useful to treat collections of items as sets upon which standard set theory
operations (like membership, subset, intersection, and union) can be applied. A convenient
representation of a set is as a list in which repeated items are not allowed and the order of the
items does not matter.

Suppose you have sets S1 and S2 represented as linked-lists, and access to the standard
primitive list operators first, rest and isEmpty. Write a recursive procedure
member(x,S1) that returns true if item x is in set S1, and false if it is not.

Provide an argument that leads to the average-case and worst-case time complexities of your
member(x,S1) algorithm in Big O notation.

Now write a recursive procedure subset(S1,S2) that returns true if set S1 is a subset of
set S2, and false if it is not. It is only allowed to call the standard primitive list operators
first, rest and isEmpty and your member procedure.

Finally, write a recursive procedure union(S1,S2) that returns the union of sets S1 and
S2 represented as linked-lists. It is only allowed to call the standard primitive list operators
makelist, first, rest and isEmpty and your member procedure.

Question 2 (14 marks)

Draw the binary search tree that results from inserting the items [19, 30, 36, 10, 40, 25, 33] in
that order into an initially empty tree.

Show how the tree rotation approach in the Lecture Notes (Section 7.10) can be used to
balance that tree.

Draw tree that results from deleting the item 30 from your balanced tree using the delete
algorithm in the Lecture Notes (Section 7.7).

Question 3 (16 marks)

In the Lecture Notes (Section 7.8), a simple procedure isbst(t) was defined that returns
true if t is a binary search tree and false if it is not. Explain why that algorithm is not
efficient, and illustrate your explanation with a simple example.

Derive an expression for the number of value comparisons C(h) this algorithm requires for a
full binary search tree of height h? [Hint: It is easiest to write C(h) as the sum of the number
of comparisons at each level, and then sum that series using the result from the lectures that
the number of nodes n in a full tree of height h is its size s(h) = 20 + 21 + 22 +… + 2h = 2h+1 –1.
To get the individual terms in the sum, think how many nodes there are at each level i in the
tree, and how many nodes further down the tree each of them is compared to.]

Deduce the time complexity of this algorithm in terms of the size n of the tree. [Hint: Derive
approximate expressions for the size n and complexity C as functions of large h, and hence
obtain an approximate expression for the complexity C as functions of large n.]

Question 4 (16 marks)

Suppose you already have an O(n) procedure tree2list(t) that takes a binary tree t
and returns a linked list of node values in such a way that they will be in ascending order if t
is a binary search tree (e.g., using an approach similar to that described in the Lecture Notes
Section 7.9). Use that procedure, and any of the standard primitive list and tree operators, to
write a procedure isbst2(t) that performs the same task as isbst(t) but more
efficiently. [Hint: You may find it most straightforward to make isbst2(t) a non-
recursive procedure that calls a separate recursive procedure.]

What are the average-case and worst-case time complexities of your improved procedure?

What scope is there to develop even more efficient procedures?

Question 5 (12 marks)

The 15 numbers {1, 2, 3, …, 15} can be stored in a perfectly balanced binary search tree of
height 3. Draw that tree.

Suppose the determination of “bigger, equal or smaller” is counted as one “comparison”. For
each of the above 15 numbers, state how many comparisons are required to determine
whether that number is in the search tree.

Compute the total number of comparisons C required to search for all 15 numbers, and hence
find the average number of comparisons A required to search for one number.

Question 6 (20 marks)

Derive expressions for C(h) and A(h) which generalize the total and average number of
comparisons computed in Question 5 to the case of a full binary search tree of any height h.
[Hint: This question is quite challenging! If your mathematics is strong, it is possible to
write C(h) as the sum of the number of comparisons at each level, and then simplify that
using the usual algebraic tricks for summing series. Or, you may find it easier to proceed in
the same way as in Lecture Notes Section 6.5 where an expression was derived for the tree
size s(h), i.e. the total number of nodes in a full binary tree of height h. First think about how
C(h+1) is related to C(h) by adding in the extra row h+1, then think about how the height h+1
tree’s C(h+1) is related to its two sub-tree’s C(h), and finally combine the two equations you
get for C(h+1) to give an equation for C(h). You can use the expression for s(h) from the
Lecture Notes or Question 3. Whichever approach you use, show all the steps involved.]

What can you deduce about the average number of comparisons required, and hence the
average time complexity of the search, for large trees?