$\newcommand{\nfrac}[2]{\frac{\displaystyle{#1}}{\displaystyle{#2}}}$
Exercises
5.1.3
Hints
- you can use floors and ceils as subroutines.
Solution
(a)
def divConqPower(a,n)
if n = 1
return a
return divConqPower(a, floor(n/2)) * divConqPower(a, ceil(n/2))
(b)
Time of basic operations $T(n)$ are $T(n) = T( \lfloor n/2 \rfloor ) + T( \lceil n/2 \rceil) + 1$. Assuming $n = 2^k$, We get $T(n) = 2T(n/2) + 1$. By master theorem, $T(n) = \Theta(n)$.
For general cases of $n$, Observe $n = 2^{\log n} \leq 2^{\lceil \log n \rceil}$, So by master theorem $T(n) \leq \mathcal{O}(2^{\lceil \log n \rceil}) \leq \mathcal{O}(2^{\log n + 1}) = \mathcal{O}(n)$. Similarly $T(n) \geq \Omega(2^{\lfloor \log n \rfloor}) = \Omega(n)$. That concludes $T(n) = \Theta(n)$ for any $n$.
Complexity proof by substitution. Recurrence is \begin{aligned} T(1) &= 0 \\ T(n) &= 2T(n/2) + 1 \end{aligned} It follows \begin{aligned} T(n) &= 2T(n/2) + 1 \\ &= 2^2 T(n/2^2) + 2^1 + 2^0 \\ &= 2^3 T(n/2^3) + 2^2 + 2^1 + 2^0 \\ &= 2^k T(n/2^k) + 2^{k-1} + \dots + 2^0 \\ &= 2^{\log n} T(n/2^{\log n}) + 2^{(\log n) - 1} + \dots + 2^0 \\ &= 2^{\log n} T(1) + 2^{(\log n) - 1} + \dots + 2^0 \\ &= 2^{(\log n) - 1} + \dots + 2^0 \\ &= 2^{\log n } - 1 = n - 1 = \Theta(n) \end{aligned}
P.S. Generally speaking we can ignore floors and ceilings in asymptotic notation (see page 885 in MIT’s Math for CS).
(c). Homework
5.1.10
Homework
5.2.8
Homework Consider pivot to be zero.
5.2.9
Homework
5.3.1
Hints
- Recall in a binary tree, a node has at most two leaves. Apply the strategy on them
- Given the height of subtrees, What can you conclude about height of the main tree?
- What is subtrees have different heights?
Solution
Same as solution manual:
Analysis. We know each node is going to count 1 operation. If we assumed total number of nodes to be $n$, then $T(n) = \Theta(n)$. If we assumed like the book the total number of internal nodes to be $n$ and leaves to be $x$, then $x + n = 2n$, so $T(n) = \Theta(n)$.
P.S. The book considers checking whether tree is empty to be the basic operation. My intuition tells me it is the max operation.
5.3.2
Hints
- Consider the problem size to be level of a node, not number of nodes.
- Consider the base case as the tree being a single node (leaf).
- Assume you can query the tree’s root, and its children.
- Recall a binary tree has at most two children for each node. Given counts of both, what can we conclude?
Solution
Commentary on the given algorithm in the question. It is flawed. Given the solution of an empty tree, we reach a flawed claim about the tree of size 1 node. Remarkably we shouldn’t consider problem size to be the number of nodes, but rather the height.
Correct algorithms:
# input: tree T with access to its root
# output: count of leaves
def leafCounter (Tree T)
# base case: the tree is empty
if T is empty
return 0
# base case: the tree is a leaf node
if T.root.children == []
return 1
# recursive step
return leafCounter(T_left) + leafCounter(T_right)
Analysis. We count one summation operation for each non-leaf node. Following the notation given in the book, $n$ as number of non-leaf nodes and $x$ as leaves, We get $T(n) = n$.
Alternatively we can consider $n$ to be total number of nodes. But we know $n = 2i+1$ where $i$ is number of non-leaf nodes. So $T(n) \approx n/2 = \Theta(n)$.
5.4.3
Homework Proving exponent rules is not germane to the course.
5.4.8
Homework Uses geometric series.
5.5.1
Hints
- Among pairs of right and left subsets, What are the least-distance ones?
- Use the given sorted property to deduce the least-distance pair.
Solution
Like the solution manual.
5.5.9
Homework