Chapter 2

Divide-and-Conquer
Divide-and-Conquer

The most-well known algorithm design strategy:

1. Divide instance of problem into two or more smaller instances

2. Solve smaller instances recursively

3. Obtain solution to original (larger) instance by combining these

solutions
Divide-and-Conquer Technique (cont.)

a problem of size n
(instance)

subproblem 1 subproblem 2
of size n/2 of size n/2

a solution to a solution to
subproblem 1 subproblem 2

a solution to It general leads to a

the original problem recursive algorithm!
 Divide-and-Conquer Examples
• Sorting: mergesort and quicksort

• Binary tree traversals

• Binary search (?)

• Multiplication of large integers

• Matrix multiplication: Strassen’s algorithm

• Closest-pair and convex-hull algorithms

General Divide-and-Conquer Recurrence

T(n) = aT(n/b) + f (n) where f(n)  (nd), d  0

Master Theorem: If a < bd, T(n)  (nd)

If a = bd, T(n)  (nd log n)
If a > bd, T(n)  (nlog b a )

Note: The same results hold with O instead of .

Examples: T(n) = 4T(n/2) + n  T(n)  ? (n^2)

(n^2log n)
T(n) = 4T(n/2) + n2  T(n)  ?
(n^3)
T(n) = 4T(n/2) + n3  T(n)  ?
Mergesort
• Split array A[0..n-1] into about equal halves and make copies of each half
in arrays B and C
• Sort arrays B and C recursively
• Merge sorted arrays B and C into array A as follows:
• Repeat the following until no elements remain in one of the
arrays:
• compare the first elements in the remaining unprocessed
portions of the arrays
• copy the smaller of the two into A, while incrementing the
index indicating the unprocessed portion of that array
• Once all elements in one of the arrays are processed, copy
the remaining unprocessed elements from the other array
into A.
Pseudocode of Mergesort
Pseudocode of Merge

Time complexity: Θ(p+q) = Θ(n) comparisons

Mergesort Example

8 3 2 9 7 1 5 4

8 3 2 9 7 1 5 4

8 3 2 9 71 5 4

8 3 2 9 7 1 5 4 The non-recursive version

of Mergesort starts from
merging single elements
3 8 2 9 1 7 4 5 into sorted pairs.

2 3 8 9 1 4 5 7

1 2 3 4 5 7 8 9
Analysis of Mergesort

T(n) = have
• All cases 2T(n/2) + Θ(n),
same T(1) = 0 Θ(n log n)
efficiency:

• Number of comparisons in the worst case is close to theoretical

minimum for comparison-based sorting:
log2 n! ≈ n log2 n - 1.44n

• Space requirement: Θ(n) (not in-place)

• Can be implemented without recursion (bottom-up)

Quicksort
• Select a pivot (partitioning element) – here, the first element
• Rearrange the list so that all the elements in the first s positions are
smaller than or equal to the pivot and all the elements in the remaining
n-s positions are larger than or equal to the pivot (see next slide for an
algorithm)

• Exchange the pivot with the last element in the first (i.e., ) subarray —
A[i]p A[i]p
the pivot is now in its final position
• Sort the two subarrays recursively
Partitioning Algorithm

or i > r
or j = l
<

Time complexity: Θ(r-l) comparisons

Quicksort Example

5 3 1 9 8 2 4 7
2 3 1 4 5 8 9 7
1 2 3 4 5 7 8 9
1 2 3 4 5 7 8 9
1 2 3 4 5 7 8 9
1 2 3 4 5 7 8 9
Analysis of Quicksort
• Best case: split in the middle — Θ(n log n)
• Worst case: sorted array! — Θ(n2) T(n) = T(n-1) + Θ(n)
• Average case: random arrays — Θ(n log n)

• Improvements:
• better pivot selection: median of three partitioning
• switch to insertion sort on small subfiles
• elimination of recursion
These combine to 20-25% improvement

• Considered the method of choice for internal sorting of large files (n ≥

10000)
Binary Search
Very efficient algorithm for searching in sorted array:
K
vs
A[0] . . . A[m] . . . A[n-1]
If K = A[m], stop (successful search); otherwise, continue
searching by the same method in A[0..m-1] if K < A[m]
and in A[m+1..n-1] if K > A[m]

l  0; r  n-1
while l  r do
m  (l+r)/2
if K = A[m] return m
else if K < A[m] r  m-1
else l  m+1
return -1
Analysis of
• Time efficiency
Binary Search
• worst-case recurrence: Cw (n) = 1 + Cw( n/2 ), Cw (1) = 1
solution: Cw(n) = log2(n+1)

This is VERY fast: e.g., Cw(106) = 20

• Optimal for searching a sorted array

• Limitations: must be a sorted array (not linked list)

• Bad (degenerate) example of divide-and-conquer

because only one of the sub-instances is solved

• Has a continuous counterpart called bisection method for solving equations in

one unknown f(x) = 0 (see Sec. 12.4)
Binary Tree Algorithms
Binary tree is a divide-and-conquer ready structure!

Ex. 1: Classic traversals (preorder, inorder, postorder)

Algorithm Inorder(T)
if T   a a
Inorder(Tleft) b c b c
print(root of T) d e   d e
Inorder(Tright) 

Efficiency: Θ(n). Why?

Each node is visited/printed once.
Binary Tree Algorithms (cont.)

Ex. 2: Computing the height of a binary tree

TL TR

h(T) = max{h(TL), h(TR)} + 1 if T   and h() = -1

Efficiency: Θ(n). Why?

Multiplication of Large Integers
Consider the problem of multiplying two (large) n-digit integers represented
by arrays of their digits such as:

A = 12345678901357986429 B = 87654321284820912836

The grade-school algorithm:

a1 a2 … an
b1 b2 … bn
(d10) d11d12 … d1n
(d20) d21d22 … d2n
…………………
(dn0) dn1dn2 … dnn

Efficiency: Θ(n2) single-digit multiplications

First Divide-and-Conquer Algorithm
A small example: A  B where A = 2135 and B = 4014
A = (21·102 + 35), B = (40 ·102 + 14)
So, A  B = (21 ·102 + 35)  (40 ·102 + 14)
= 21  40 ·104 + (21  14 + 35  40) ·102 + 35  14

In general, if A = A1A2 and B = B1B2 (where A and B are n-digit,

A1, A2, B1, B2 are n/2-digit numbers),
A  B = A1  B1·10n + (A1  B2 + A2  B1) ·10n/2 + A2  B2

Recurrence for the number of one-digit multiplications M(n):

M(n) = 4M(n/2), M(1) = 1
Solution: M(n) = n2
Second Divide-and-Conquer Algorithm

A  B = A1  B1·10n + (A1  B2 + A2  B1) ·10n/2 + A2  B2

The idea is to decrease the number of multiplications from 4 to 3:

(A1 + A2 )  (B1 + B2 ) = A1  B1 + (A1  B2 + A2  B1) + A2  B2,

I.e., (A1  B2 + A2  B1) = (A1 + A2 )  (B1 + B2 ) - A1  B1 - A2  B2,
which requires only 3 multiplications at the expense of (4-1) extra add/sub.

Recurrence for the number of multiplications M(n):

M(n) = 3M(n/2), M(1) = 1
Solution: M(n) = 3log 2n = nlog 23 ≈ n1.585
What if we count
both multiplications
and additions?
Example of Large-Integer Multiplication

2135  4014

= (21*10^2 + 35) * (40*10^2 + 14)

= (21*40)*10^4 + c1*10^2 + 35*14
where c1 = (21+35)*(40+14) - 21*40 - 35*14, and
21*40 = (2*10 + 1) * (4*10 + 0)
= (2*4)*10^2 + c2*10 + 1*0
where c2 = (2+1)*(4+0) - 2*4 - 1*0, etc.

This process requires 9 digit multiplications as opposed to 16.

Conventional Matrix Multiplication

• Brute-force algorithm
c00 c01 a00 a01 b00 b01
= *
c10 c11 a10 a11 b10 b11

a00 * b00 + a01 * b10 a00 * b01 + a01 * b11

=
a10 * b00 + a11 * b10 a10 * b01 + a11 * b11

8 multiplications Efficiency class in general:  (n3)

4 additions
 Strassen’s Matrix Multiplication
• Strassen’s algorithm for two 2x2 matrices (1969):
c00 c01 a00 a01 b00 b01
= *
c10 c11 a10 a11 b10 b11

m1 + m4 - m5 + m7 m3 + m5
=
m2 + m4 m1 + m3 - m2 + m6
• m1 = (a00 + a11) * (b00 + b11)
• m2 = (a10 + a11) * b00
• m3 = a00 * (b01 - b11)
• m4 = a11 * (b10 - b00)
• m5 = (a00 + a01) * b11
• m6 = (a10 - a00) * (b00 + b01)
7 multiplications
• m7 = (a01 - a11) * (b10 + b11)
18 additions
 Strassen’s Matrix Multiplication
Strassen observed [1969] that the product of two matrices can be
computed in general as follows:

C00 C01 A00 A01 B00 B01

= *
C10 C11 A10 A11 B10 B11

M1 + M4 - M5 + M7 M3 + M5
=
M2 + M4 M1 + M3 - M2 + M6
Formulas for Strassen’s Algorithm
M1 = (A00 + A11)  (B00 + B11)

M2 = (A10 + A11)  B00

M3 = A00  (B01 - B11)

M4 = A11  (B10 - B00)

M5 = (A00 + A01)  B11

M6 = (A10 - A00)  (B00 + B01)

M7 = (A01 - A11)  (B10 + B11)

 Analysis of Strassen’s Algorithm
If n is not a power of 2, matrices can be padded with zeros.

What if we count both

Number of multiplications: multiplications and additions?
M(n) = 7M(n/2), M(1) = 1
Solution: M(n) = 7log 2n = nlog 27 ≈ n2.807 vs. n3 of brute-force alg.

Algorithms with better asymptotic efficiency are known but they

are even more complex and not used in practice.
 Closest-Pair Problem by Divide-and-Conquer
Step 0 Sort the points by x (list one) and then by y (list two).

Step 1 Divide the points given into two subsets S1 and S2 by a vertical line x =
c so that half the points lie to the left or on the line and half the points
lie to the right or on the line.
Closest Pair by Divide-and-Conquer (cont.)

Step 2 Find recursively the closest pairs for the left and right
subsets.

Step 3 Set d = min{d1, d2}

We can limit our attention to the points in the symmetric
vertical strip of width 2d as possible closest pair. Let C1
and C2 be the subsets of points in the left subset S1 and of
the right subset S2, respectively, that lie in this vertical
strip. The points in C1 and C2 are stored in increasing
order of their y coordinates, taken from the second list.

Step 4 For every point P(x,y) in C1, we inspect points in

C2 that may be closer to P than d. There can be no more
than 6 such points (because d ≤ d2)!
Closest Pair by Divide-and-Conquer: Worst Case

The worst case scenario is depicted below:

Efficiency of the Closest-Pair Algorithm

Running time of the algorithm (without sorting) is:

T(n) = 2T(n/2) + M(n), where M(n)  Θ(n)

By the Master Theorem (with a = 2, b = 2, d = 1)

T(n)  Θ(n log n)

So the total time is Θ(n log n).

Quickhull Algorithm
Convex hull: smallest convex set that includes given points. An O(n^3)
bruteforce time is given in Levitin, Ch 3.
• Assume points are sorted by x-coordinate values
• Identify extreme points P1 and P2 (leftmost and rightmost)
• Compute upper hull recursively:
• find point Pmax that is farthest away from line P1P2
• compute the upper hull of the points to the left of line P1Pmax
• compute the upper hull of the points to the left of line PmaxP2
• Compute lower hull in a similar manner
Pmax

P2

P1
Efficiency of Quickhull Algorithm
• Finding point farthest away from line P1P2 can be done in linear time
• Time efficiency: T(n) = T(x) + T(y) + T(z) + T(v) + O(n),
where x + y + z +v <= n.
• worst case: Θ(n2) T(n) = T(n-1) + O(n)
• average case: Θ(n) (under reasonable assumptions about
distribution of points given)

• If points are not initially sorted by x-coordinate value, this can be

accomplished in O(n log n) time

• Several O(n log n) algorithms for convex hull are known