CHAPTER 5 PRIORITY QUEUES (HEAPS)
description
Transcript of CHAPTER 5 PRIORITY QUEUES (HEAPS)
CHAPTER 5
PRIORITY QUEUES (HEAPS)
§1 ADT Model
Objects: A finite ordered list with zero or more elements.
Operations:
PriorityQueue Initialize( int MaxElements );
void Insert( ElementType X, PriorityQueue H );
ElementType DeleteMin( PriorityQueue H );
ElementType FindMin( PriorityQueue H );
—— delete the element with the highest \ lowest priority
1/15
§2 Simple Implementations Array :
Insertion — add one item at the end ~ ( 1 )Deletion — find the largest \ smallest key ~ ( n ) remove the item and shift array ~ O( n )
Linked List :
Insertion — add to the front of the chain ~ ( 1 )Deletion — find the largest \ smallest key ~ ( n ) remove the item ~ ( 1 )
Ordered Array :
Insertion — find the proper position ~ O( n ) shift array and add the item ~ O( n )Deletion — remove the first \ last item ~ ( 1 )
Ordered Linked List :
Insertion — find the proper position ~ O( n ) add the item ~ ( 1 )Deletion — remove the first \ last item ~ ( 1 )
Better since there are never more deletions than insertions
2/15
Binary Search Tree :
§2 Simple Implementations
Ah! That’s a good idea! Both insertion and deletion will take
O(log N) only.
Well, insertions are random, but deletions are NOT. We are supposed to delete
The minimum element only.
Oh, right, then we must always delete from the left subtrees….
But hey, what if we keep a balanced tree?
Hey you are getting smarter! Yes a balanced tree such as AVL tree
is not a bad idea since only a constant factor will be added to
the run time. However…
Oh no… what’s wrong?There are many operations related to AVL tree that we don’t really
need for a priority queue.Besides, pointers are
always dangerous.
I bet you have a better option?Now you begin to know me
3/15
§3 Binary Heap1. Structure Property:
【 Definition】 A binary tree with n nodes and height h is complete iff its nodes correspond to the nodes numbered from 1 to n in the perfect binary tree of height h.
4
8 9
5
10 11
6
12 13
7
14 15
2 3
1
A complete binary tree of height h has between
and nodes.
2h
2h+1 1 h = log N
Array Representation : BT [ n + 1 ] ( BT [ 0 ] is not used)
D
H I
E
J
F G
B C
ABT 0 1
A2B
3C
4D
5E
6F
7G
8H
9I
10J
11 12 13
4/15
§3 Binary Heap
【 Lemma】 If a complete binary tree with n nodes is represented sequentially, then for any node with index i, 1 i n, we have:
ni
niiichildright
ni
niiichildleft
i
iiiparent
12 ifNone
12 if12)(_ ofindex (3)
2 ifNone
2 if2)(_ ofindex (2)
1 ifNone
1 if2)( ofindex (1)
5/15
§3 Binary Heap
PriorityQueue Initialize( int MaxElements ) { PriorityQueue H; if ( MaxElements < MinPQSize )
return Error( "Priority queue size is too small" ); H = malloc( sizeof ( struct HeapStruct ) ); if ( H ==NULL )
return FatalError( "Out of space!!!" ); /* Allocate the array plus one extra for sentinel */ H->Elements = malloc(( MaxElements + 1 ) * sizeof( ElementType )); if ( H->Elements == NULL )
return FatalError( "Out of space!!!" ); H->Capacity = MaxElements; H->Size = 0; H->Elements[ 0 ] = MinData; /* set the sentinel */ return H; }
6/15
§3 Binary Heap2. Heap Order Property:
【 Definition】 A min tree is a tree in which the key value in each node is no larger than the key values in its children (if any). A min heap is a complete binary tree that is also a min tree.
Note: Analogously, we can declare a max heap by changing the heap order property.
Note: Analogously, we can declare a max heap by changing the heap order property.
9
6
5
3
[1]
[2] [3]
[4]
A max heap
10
20
50
83
[1]
[2] [3]
[4]
A min heap
The largest key The smallest key
7/15
§3 Binary Heap3. Basic Heap Operations:
insertion
10
12
15
20
[1]
[2] [3]
[4]
18[5] [6]
Sketch of the idea:
The only possible position for a new node
since a heap must be a complete binary tree.
Case 1 : new_item = 21
21
20 21<
Case 2 : new_item = 17
17
20 17>
17
20
10 17<
Case 3 : new_item = 9 20 9>
9
10 9>
9
10
8/15
§3 Binary Heap
/* H->Element[ 0 ] is a sentinel */ void Insert( ElementType X, PriorityQueue H ) { int i;
if ( IsFull( H ) ) { Error( "Priority queue is full" ); return;
}
for ( i = ++H->Size; H->Elements[ i / 2 ] > X; i /= 2 ) H->Elements[ i ] = H->Elements[ i / 2 ];
H->Elements[ i ] = X; }
Percolate up
Faster than swap
H->Element[ 0 ] is a sentinel that is no larger
than the minimum element in the heap.
T (N) = O ( log N )
9/15
§3 Binary Heap
DeleteMin
Sketch of the idea:
10
12
15
20
[1]
[2] [3]
[4]
18[5]
The node which must beremoved to keep a
complete binary tree.
move 18 up to the root18
find the smaller child of 18 12 18<18
12
15 18<18
15
Ah! That’s simple --we only have to delete
the root node ...And re-arrange
the rest of the tree so that it’s still a min heap.
T (N) = O ( log N )
10/15
§3 Binary Heap
ElementType DeleteMin( PriorityQueue H ) { int i, Child; ElementType MinElement, LastElement; if ( IsEmpty( H ) ) { Error( "Priority queue is empty" ); return H->Elements[ 0 ]; } MinElement = H->Elements[ 1 ]; /* save the min element */ LastElement = H->Elements[ H->Size-- ]; /* take last and reset size */ for ( i = 1; i * 2 <= H->Size; i = Child ) { /* Find smaller child */ Child = i * 2; if (Child != H->Size && H->Elements[Child+1] < H->Elements[Child])
Child++; if ( LastElement > H->Elements[ Child ] ) /* Percolate one level */
H->Elements[ i ] = H->Elements[ Child ]; else break; /* find the proper position */ } H->Elements[ i ] = LastElement; return MinElement; }
Percolate down
What if this condition is
omitted?
Can we remove it by adding another
sentinel?
11/15
§3 Binary Heap4. Other Heap Operations:
Note: Finding any key except the minimum one will have to take a linear scan through the entire heap.
Note: Finding any key except the minimum one will have to take a linear scan through the entire heap.
DecreaseKey ( P, , H )
Lower the value of the key in the heap H at position P by a positive amount of ……so my programs can run with highest priority .
sys. admin.
IncreaseKey ( P, , H )
Percolate upPercolate up
sys. admin.
Increases the value of the key in the heap H at position P by a positive amount of ……drop the priority of a process that is consuming excessive CPU time.
Percolate downPercolate down
12/15
§3 Binary Heap
Delete ( P, H )
Remove the node at position P from the heap H …… delete the process that is terminated (abnormally) by a user.
sys. admin.
DecreaseKey(P, , H); DeleteMin(H)DecreaseKey(P, , H); DeleteMin(H)
BuildHeap ( H )
Place N input keys into an empty heap H.
sys. admin.
N successive Insertions ? N successive Insertions ? Nehhhhh that would be
toooo slow !
30
100 20
10
90 60
70
50 120
110
140 130
80 40
150
150, 80, 40, 30, 10, 70, 110, 100, 20, 90, 60, 50, 120, 140, 130
PercolateDown (7)PercolateDown (6)
50
70
PercolateDown (5)PercolateDown (4)
20
30
PercolateDown (3)PercolateDown (2)
80
10
80
60
PercolateDown (1)
150
10
150
20
150
30
T (N) = ?
13/15
§3 Binary Heap
【 Theorem】 For the perfect binary tree of height h containing 2h+1 1 nodes, the sum of the heights of the nodes is 2h+1 1 (h + 1).
T ( N ) = O ( N )
§4 Applications of Priority Queues
〖 Example〗 Given a list of N elements and an integer k. Find the kth largest element.
How many methods can you think of to solve this problem? What are
their complexities?
Home work:p.180 5.2, 5.3, 5.4
A test case for operationson a heap
14/15
§5 d-Heaps ---- All nodes have d children
3
1513 6
5
178 9
2
74 10
911
1
3-heap
Question: Shall we make d as large as possible?
Note: DeleteMin will take d 1 comparisons to find the smallest child. Hence the total time complexity would be O(d logd N).
*2 or /2 is merely a bit shift, but *d or /d is not. When the priority queue is too large to fit entirely in main
memory, a d-heap will become interesting.
Note: DeleteMin will take d 1 comparisons to find the smallest child. Hence the total time complexity would be O(d logd N).
*2 or /2 is merely a bit shift, but *d or /d is not. When the priority queue is too large to fit entirely in main
memory, a d-heap will become interesting.
Home work:p.181 5.13
Index computations for parents and children
in a d-heap
15/15