Download presentation
Presentation is loading. Please wait.
Published byDiego Worrick Modified over 9 years ago
1
Chapter 7 Sorting Part II
2
7.3 QUICK SORT
3
Example 4151264 left right pivot i j 5 > pivot and should go to the other side. 2 < pivot and should go to the other side. Interchange a[i] and a[j] 25 Stop, because i == j. a[j] will eventually stop at a position where a[j] < pivot. Interchange a[j] and pivot. 144 Stop; because i > j.
4
Algorithm void QuickSort(int a[], int left, int right) { if (left < right) { pivot = a[left]; i = left; j = right+1; while (i < j) { for (i++; i<j and a[i] < pivot; i++) ; for (j--; i = pivot; j--) ; if (i < j) interchages a[i] and [i]; } interchange a[j] and pivot; QuickSort(a, left, j-1); QuickSort(a, j+1 right;); } void QuickSort(int a[], int left, int right) { if (left < right) { pivot = a[left]; i = left; j = right+1; while (i < j) { for (i++; i<j and a[i] < pivot; i++) ; for (j--; i = pivot; j--) ; if (i < j) interchages a[i] and [i]; } interchange a[j] and pivot; QuickSort(a, left, j-1); QuickSort(a, j+1 right;); }
5
Analysis of QuickSort() Worst case: ◦ Consider a list is stored. The smallest one is always chosen as pivot. In 1st iteration, n-1 elements are examined. In second iterations, n-2 elements are examined... Totally, the execution steps are n-1 + n-2 + … + 1 = O(n 2 ) The time complexity is O(n 2 ); 16425
6
Lemma 7.1 Let T avg (n) be the expect time for function QuickSort() to sort a list with n records. Then there exists a constant k such that T avg (n) ≦ knlog e n for n ≧ 2. ◦ In other words, T avg (n) = O(nlogn)
7
Variations The position of the pivot decides the time complexity of QuickSort(). The best choice for the pivot is the median. ◦ Variations: Median-of-three: select the median among three records: the most left, the most right, and the middle one. Random: select the pivot randomly.
8
7.4 HOW FAST CAN WE SORT? (DECISION TREE)
9
Consideration What is the best computing time for sorting that we can hope for? ◦ Suppose the only operations permitted on keys are comparisons and exchanges. In this section, we shall prove O(n logn) is the best possible time. Using decision tree.
10
Example 7.4 Decision tree for Insertion Sort working on [K 1, K 2, K 3 ] K1 ≦K2K1 ≦K2 [1, 2, 3] K2 ≦K3K2 ≦K3 Stop [1, 2, 3] K1 ≦K3K1 ≦K3 [1, 3, 2] Stop [3, 1, 2][1, 3, 2] K1 ≦K3K1 ≦K3 [2, 1, 3] Stop K2 ≦K3K2 ≦K3 [2, 3, 1] Stop [3, 2, 1][2, 3, 1] [2, 1, 3] Y Y Y Y Y N N N N N I IIIII IV VVI
11
Observations The leaf nodes denote the states to terminate. The number of permutations is 3! = 6. ◦ n! possible permutation for n records to sort. A path from the root to some leaf node represents one of n! possibilities. The maximum depth of the tree is 3. ◦ The depth represents the number of comparisons.
12
Theorem 7.1 Any decision tree that sorts n distinct elements has a height of at least log 2 (n!)+1. ◦ When sorting n elements, there are n! different possible results. Every decision tree for the sorting must at least have n! leaves. ◦ A decision tree is a binary tree; therefore 2 k-1 leaves if its height is k.
13
Corollary Any algorithm that sorts only by comparisons must have a worst case computing time of Ω (n logn). Proof ◦ By theorem, there is a path of length log 2 n! So,
14
7.5 MERGE SORT
15
Merging Consider how to merge two ordered lists. initList mergeList l mm+1n sorted merge
16
Example 45711236814 2 2 3 3 4 4 5 5 6 6 7 7 8 8 11 14 l mm+1n i1i2i1 iResult i1 i2 iResult iResult = l initList mergeList i1i2 iResult Copy the small one to mergeList Copy the rest to mergeList
17
void Merge(int *initList, int *mergeList, int l, int m, int n) { int i1=l, iResult=l, i2=m+1; while (i1 <= m && i2 <= n) { if (initList[i1] <= initList[i2]) { mergeList[iResult++] = initList[i1]; i1++; } else { mergeList[iResult++] = initList[i2]; i2++; } for (i1; i1<=m; i1++) mergeList[iResult++] = initList[i1]; for (i2; i2<=n; i2++) mergeList[iResult++] = initList[i2]; }
18
Analysis of Merge() Time complexity: ◦ The while-loop and two for-loops examine each element in initList exactly once. ◦ The time complexity is O(n- l +1). Space complexity: ◦ The additional array mergeList is required to store the merged result. ◦ Space complexity is O(n- l +1).
19
7.5.2 Iterative Merge Sort L: the maximum number of records in a block. 265771611159154819 5 261 7711 6115 5919 48 1 5 26 7711 15 59 6119 48 1 5 11 15 26 56 61 7719 48 1 5 11 15 19 26 48 56 61 77 L = 1 L = 2 L = 4 L = 8 L = 16
20
Merging blocks with length of L: 5 261 7711 6115 5919 48 L = 2 ii+L-1i+Li+2L-1i+2L 0123
21
Adjacent pairs of blocks of size L are merged from initList to resultList. n is the number of records in initList. void MergePass(int *initList, int *resultList, int n, int L) { int i; for (i=0; i<=n-2*L; i+=2*L) Merge(initList, resultList, i i+L-1, i+2L-1); if (i + L - 1 < n-1) Merge(initList, resultList, i, i+L-1, n-1); else { for (i; i<n; i++) resultList[i] = resultList[i]; } void MergePass(int *initList, int *resultList, int n, int L) { int i; for (i=0; i<=n-2*L; i+=2*L) Merge(initList, resultList, i i+L-1, i+2L-1); if (i + L - 1 < n-1) Merge(initList, resultList, i, i+L-1, n-1); else { for (i; i<n; i++) resultList[i] = resultList[i]; } Merge remaining blocks of L < size < 2L
22
01234567 i<=n-2L. Merge adjacent blacks. i<=n-2L. Merge adjacent blacks. 012346 i+L-1 < n-1. Merge from i to n-1. i+L-1 < n-1. Merge from i to n-1. 5 01234 i+L-1 >= n-1. Copy the rest. i+L-1 >= n-1. Copy the rest. 5
23
Merge Sort L denotes the length of block currently being merged. void MergeSort(int *a, int n) { int *tempList = new int[n]; for (int L=1; L<n; L*=2) { MergePass(a, tempList, n, L); L *= 2; MergePass(tempList, a, n, L); } delete [] tempList; } The result is put into tempList Do the next pass directly. Merge records from tempList to a
24
Analysis of MergeSort() Suppose there are n records. Space complexity: O(n). Time Complexity ◦ MergePass(): O(n). ◦ MergeSort(): A total passes are made over the data. Therefore, the time complexity O(nlogn). LL n
25
7.5.2 Recursive Merge Sort We divide the list into two roughly equal parts and sort them recursively. sorted merge leftright
26
265771611159154819 265771611159154819 265771611159154819 2651159 5 26 5 26 77 11 59 48 191 6111 15 59 1 5 26 61 7711 15 19 48 59 1 5 11 15 19 26 48 59 61 77 7715 26511594819161
27
s void Merge(int *initList, int s, int m, int e) { int *temp = new int[e-s+1]; int i1=s, iResult=0, i2=m+1; while (i1 <= m && i2 <= e) { if (initList[i1] <= initList[i2]) { temp[iResult++] = initList[i1]; i1++; } else { temp[iResult++] = initList[i2]; i2++; } for (i1; i1<=m; i1++) temp[iResult++] = initList[i1]; for (i2; i2<=n; i2++) temp[iResult++] = initList[i2]; for (i=0; i<iResult; i++) initList[i] = temp[i]; delete [] temp; }
28
MergeSort() start and end respectively denote the left end and right end to be sorted in the array a. void MergeSort(int *a, int start, int end) { if (end <= start) return; middle = (start + end) / 2; MergeSort(a, start, middle); MergeSort(a, middle+1, end); Merge(a, start, middle, end); }
29
Analysis of Recursive Merge Sort Suppose there are n records to be sorted. Time complexity O(nlogn)
30
Variation Natural Merge Sort ◦ Make an initial pass over the data To determine the sublists of records that are in order.
31
7.6 HEAP SORT
32
Discussion Merge Sort ◦ In worst case and average case, the time complexity is O(nlogn). ◦ However, additional storage is required. There is O(1) space merge algorithm, but it is much slower than the original one. Heap Sort ◦ Only a fixed amount of additional storage is required. ◦ The time complexity is also O(nlogn). ◦ Using max heap.
33
Selection Sort Suppose there are n records in the list. How to sort the list? ◦ First, find the largest and put it at the position n-1. ◦ Second, find the largest from 0 to n-2 (the second largest), and put it at the position n-2. … ◦ In the i-th iteration, the i-th largest is selected and is put at the position of n-i. Consider how to sort [3, 4, 1, 5, 2].
34
34152 Select 5. 34125 Select 4. 32145 34125 Select 3. 12345 32145 Select 2. 1234512345 Select 1. 1234512345
35
Analysis of Selection In the i-th iteration, O(n-i+1) computing time is required to select the i-th largest. O(n) + O(n-1) + … + O(1) = O(n 2 ). ◦ How do we decrease the time complexity? Improve the approach of selecting the maximum. Using max heap The deletion of the maximum from a max heap is O(log n), when there are n elements in the heap. Note: when using heap sort, the data is stored in [1;n].
36
Example 26 [1] 5 [2] 77 [3] 1 [4] 61 [5] 11 [6] 59 [7] 15 [8] 48 [9] 19 [10] 77 [1] 61 [2] 59 [3] 48 [4] 19 [5] 11 [6] 26 [7] 15 [8] 1 [9] 5 [10] Initial Array Initial Heap
37
61 [1] 48 [2] 59 [3] 15 [4] 19 [5] 11 [6] 26 [7] 5 [8] 1 [9] 59 [1] 48 [2] 26 [3] 15 [4] 19 [5] 11 [6] 1 [7] 5 [8] Heap Size = 9 Sorted = [77] Heap Size = 8 Sorted = [61, 77]
38
48 [1] 19 [2] 26 [3] 15 [4] 5 [5] 11 [6] 1 [7] 26 [1] 19 [2] 11 [3] 15 [4] 5 [5] 1 [6] Heap Size = 7 Sorted = [59, 61, 77] Heap Size = 69 Sorted = [48, 59, 61, 77]
39
Adjust() Adjust binary tree with root to satisfy heap property. void Adjust(int *a, int root, int n) { int e = a[root]; for (int j=2*root; j<=n; j*=2) { if (j < n && a[j] < a[j+1]) //j is max child of its parent j++; if (e >= a[j]) break; a[j/2] = a[j]; } a[j/2] = e; }
40
Heap Sort void HeapSort(int *a, int n) { for (int i=n/2; i>=1; i--) //heapify Adjust(a, i, n); for (int i=n-1; i>=1; i--) //sort { swap(a[1], a[i+1]); Adjust(a, 1, i); }
41
Analysis of HeapSort() Space complexity: O(1). Time complexity: ◦ Suppose the tree has k levels. The number of nodes on level i is ≦ 2 i-1. ◦ In the first loop, Adjust() is called once for each node that has a child.
42
Analysis of HeapSort() Space complexity: O(1). Time complexity: ◦ Suppose the tree has k levels. The number of nodes on level i is ≦ 2 i-1. ◦ In the first loop, Adjust() is called once for each node that has a child. Time complexity:
43
Analysis of HeapSort() ◦ In the next loop, n-1 times of Adjust() are made with maximum depth k = and swap is invoked n-1 times. ◦ Consequently, time computing time for the loop is O(n logn). Overall, the time complexity for HeapSort() is O(n logn).
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.