CS 373: Design & Analysis of

Algorithms

Linear-Time Sorting Algorithms

Dr. Sifat Momen 1 10/27/2018

 Sorting So Far
 Insertion sort:
 Easy to code
 Fast on small inputs (less than ~50 elements)
 Fast on nearly-sorted inputs
 O(n2) worst case
 O(n2) average (equally-likely inputs) case
 O(n2) reverse-sorted case

Dr. Sifat Momen 1 10/27/2018

 Sorting So Far
 Merge sort:
 Divide-and-conquer:
 Split array in half
 Recursively sort subarrays
 Linear-time merge step
 O(n lg n) worst case
 Doesn’t sort in place

Dr. Sifat Momen 1 10/27/2018

 Sorting So Far
 Heap sort:
 Uses the very useful heap data structure
 Complete binary tree
 Heap property: parent key > children’s keys
 O(n lg n) worst case
 Sorts in place
 Fair amount of shuffling memory around

Dr. Sifat Momen 1 10/27/2018

 Sorting So Far
 Quick sort:
 Divide-and-conquer:
 Partition array into two subarrays, recursively sort
 All of first subarray < all of second subarray
 No merge step needed!
 O(n lg n) average case
 Fast in practice
 O(n2) worst case
 Naïve implementation: worst case on sorted input
 Address this with randomized quicksort

Dr. Sifat Momen 1 10/27/2018

 How Fast Can We Sort?
 We will provide a lower bound, then beat it
 How do you suppose we’ll beat it?
 First, an observation: all of the sorting
algorithms so far are comparison sorts
 The only operation used to gain ordering
information about a sequence is the pairwise
comparison of two elements
 Theorem: all comparison sorts are (n lg n)

Dr. Sifat Momen 1 10/27/2018

 Comparison Model
 All the items are black boxes.
 Only operations allowed are comparisons (<,
≤, >, ≥, =)
 Time cost = # of comparisons made.

Dr. Sifat Momen 1 10/27/2018

 Decision Trees
 Decision trees provide an abstraction of
comparison sorts
 A decision tree represents the comparisons made
by a comparison sort. Every thing else ignored

 What do the leaves represent?

 How many leaves must there be?

Dr. Sifat Momen 1 10/27/2018

 Decision Trees

Dr. Sifat Momen 1 10/27/2018

 Decision Trees
Decision Tree Algorithm
Every internal node Binary decision (comparison)
Leaf Answer
Root  Leaf Algorithm execution
Length of the path Running time
Height of the tree Worst case running time

Dr. Sifat Momen 1 10/27/2018

 Lower Bound For
Comparison Sorting
 What’s the minimum # of leaves?
 What’s the maximum # of leaves of a binary
tree of height h?
 Clearly the minimum # of leaves is less than or
equal to the maximum # of leaves

Dr. Sifat Momen 1 10/27/2018

 Lower Bound For
Comparison Sorting
 So we have…
n!  2h
 Taking logarithms:
lg (n!)  h
 Stirling’s approximation tells us:

n
n
n!   
e n
 Thus: n
h  lg 
e
Dr. Sifat Momen 1 10/27/2018
 Lower Bound For
Comparison Sorting
 So we have
n
n
h  lg 
e

 n lg n  n lg e

 Thus  n minimum
 the lg n  height of a decision tree is (n lg n)

Dr. Sifat Momen 1 10/27/2018

 Lower Bound For
Comparison Sorts
 Thus the time to comparison sort n elements is
(n lg n)
 Corollary: Heapsort and Mergesort are
asymptotically optimal comparison sorts
 But the name of this lecture is “Sorting in
linear time”!
 How can we do better than (n lg n)?

Dr. Sifat Momen 1 10/27/2018

 Sorting In Linear Time
 Counting sort
 No comparisons between elements!
 But…depends on assumption about the numbers
being sorted
 We assume numbers are in the range 1.. k
 The algorithm:
 Input: A[1..n], where A[j]  {1, 2, 3, …, k}
 Output: B[1..n], sorted (notice: not sorting in place)
 Also: Array C[1..k] for auxiliary storage

Dr. Sifat Momen 1 10/27/2018

 Counting Sort
1 CountingSort(A, B, k)
2 Let C[0.. K] be a new array
2 for i=0 to k
3 C[i]= 0;
4 for j=1 to A.length
5 C[A[j]] = C[A[j]] + 1;
6 for i=1 to k
7 C[i] = C[i] + C[i-1];
8 for j=A.length downto 1
9 B[C[A[j]]] = A[j];
10 C[A[j]] = C[A[j]] - 1;

Dr. Sifat Momen 1 10/27/2018

 Counting Sort
 Total time: O(n + k)
 Usually, k = O(n)
 Thus counting sort runs in O(n) time
 But sorting is (n lg n)!
 No contradiction--this is not a comparison sort (in
fact, there are no comparisons at all!)
 Notice that this algorithm is stable

Dr. Sifat Momen 1 10/27/2018

 Radix Sort
 Intuitively, you might sort on the most
significant digit, then the second msd, etc.
 Problem: lots of intermediate piles of cards
(read: scratch arrays) to keep track of
 Key idea: sort the least significant digit first

RadixSort(A, d)
for i=1 to d
StableSort(A) on digit i

Dr. Sifat Momen 1 10/27/2018

 Radix Sort

Dr. Sifat Momen 1 10/27/2018

 Radix Sort
 What sort will we use to sort on digits?
 Counting sort is obvious choice:
 Sort n numbers on digits that range from 1..k
 Time: O(n + k)
 Each pass over n numbers with d digits takes
time O(n+k), so total time O(dn+dk)
 When d is constant and k=O(n), takes O(n) time

Dr. Sifat Momen 1 10/27/2018

 Radix Sort
 In general, radix sort based on counting sort is
 Fast
 Asymptotically fast (i.e., O(n))
 Simple to code
 A good choice

Dr. Sifat Momen 1 10/27/2018

 Bucket Sort

Dr. Sifat Momen 1 10/27/2018

 Bucket Sort

Dr. Sifat Momen 1 10/27/2018

 Time Complexity

Dr. Sifat Momen 1 10/27/2018