Compression Maps

• h(k) = k mod m k: key, m: #slots

• h(k) = ⌊m(k A mod 1)⌋ 0<A<1

• h(k) = | ak + b | mod n

22
Compression Maps
• h(k) = k mod m
b
• Option 1: m = 2
h(k) gives the b least significant bits of k
All keys with the same ending go to the same place.

• Option 2: m is prime
• Ensures uniform distribution
• Avoid using prime numbers to close to a power of 2.
23
Compression Maps
• h(k) = ⌊m(k A mod 1)⌋ 0<A<1

p
• Value of m is not critical. Can be chosen as 2

• The choice of A depends on the characteristics of data.

5−1
A= corresponds to Fibonacci Hashing
2

24
Compression Maps
• h(k) = | ak + b | mod n

• a should not be a multiple of n

a and n should be co-prime

25
Universal Hashing
• Create a set of hash functions H, from which h can be randomly
selected.

• For a randomly chosen hash function h ∈ H,

1
P {h(k) = h(l)} ≤
m

26
Handling Collision
• A general class of techniques for handling collisions is Open Addressing

• All elements are stored in the hash table n ≤ m

• Inserting and searching in a hash table requires systematically probing

the available slots.

27
Open Addressing
• Probing can involve multiple attempts to find an available slot

• The hash function h is modified to include the probe number i as the

second parameter.
h : U × {0,1,…, m − 1} ⟶ {0,1,…, m − 1}

• The probe sequence for a given key k is a permutation of

⟨0,1,2,…, m − 1⟩

28
Linear Probing
• If there is a collision, the probe advances to the next slot

• Uses less memory than chaining

• Slower than chaining

29
Linear Probing . . . . . . . . . .
• h(k) = k mod 10 0 1 2 3 4 5 6 7 8 9

Insert 89 . . . . . . . . . 89
0 1 2 3 4 5 6 7 8 9
Insert 18 . . . . . . . . 18 89
0 1 2 3 4 5 6 7 8 9
Insert 49 49 . . . . . . . 18 89
0 1 2 3 4 5 6 7 8 9
Insert 58 49 58 . . . . . . 18 89
0 1 2 3 4 5 6 7 8 9
30
Linear Probing
• How to perform deletion?

. . . . . . . . . . . .
0 1 2 3 4 5 6 7 8 9 10 11

31
Double Hashing
• Use two hash functions h1(k) and h2(k)

• h1(k) is the position in the table where we first check for key k

• h2(k) is the offset to be used for moving to the next location.

• DoubleHashing_Insert(k):
if (table is full) error; probe = h1(k); offset = h2(k)
while (table[probe] is occupied)
probe = (probe + offset) mod m
table[probe] = k
32
 Double Hashing Insert 89 . . . . . . . . . 89
h1(k) = k mod 10 0 1 2 3 4 5 6 7 8 9
h2(k) = 7 − (k mod 7) Insert 18 . . . . . . . . 18 89
0 1 2 3 4 5 6 7 8 9
Insert 49 . . . . . . 49 . 18 89
0 1 2 3 4 5 6 7 8 9
Insert 58 . . . 58 . . 49 . 18 89
0 1 2 3 4 5 6 7 8 9

Insert 69 69 . . 58 . . 49 . 18 89
0 1 2 3 4 5 6 7 8 9
33
 Double Hashing Example
h1(k) = k mod 13 . . . . . 18 . . . . . . .

h2(k) = 8 − (k mod 8) 0 1 2 3 4 5 6 7 8 9 10 11 12
. . 41 . . 18 . . . . . . .
0 1 2 3 4 5 6 7 8 9 10 11 12
Insert
18 . . 41 . . 18 . . . 22 . . .
41 0 1 2 3 4 5 6 7 8 9 10 11 12
22
44 . 41 . . 48 . . . 22 . . .
44
59 0 1 2 3 4 5 6 7 8 9 10 11 12
32 44 . 41 . . 48 . 59 . 22 . . .
31
73 0 1 2 3 4 5 6 7 8 9 10 11 12
44 . 41 73 . 48 32 59 31 22 . . .
34
Analysis of Double Hashing
• The load factor λ is less than 1
• We assume that every probe looks at a random location in the table

• 1 − λ fraction of the table is empty

• Expected number of probes required to find an empty location
1
(unsuccessful search) is
1−λ

35
• Average number of probes for a successful search = average number of
probes required to insert all elements

• To insert an element we need to find an empty location

Inserting Avg # of probes Total # of probes

First m/2 <=2 m
Next m/4 <=4 m
Next m/8 <=8 m

36
 m m m m
• No of probes required to insert 2 + 4 + 8 + … + 2i elements

1
= no of probes required to leave i fraction of the table empty = m×i
2

• No of probes required to leave 1 − λ fraction of the table empty

(1 − λ)
1
= m log = − m log(1 − λ)

• Average number of probes required to insert n elements is

(n) (λ)
m 1
− log(1 − λ) = − log(1 − λ)
37
Expected number of probes
• Load factor λ < 1 for probing

• Analysis of probing uses uniform hashing assumption — any permutation is

equally likely.

unsuccessful Successful
Chaining
O(1 + λ) O(1 + λ)

(1 − λ) (λ 1 − λ)
Probing
1 1 1
O O ln
38
Trees

1
Definitions A
• A, B, C, H — internal nodes
• D, E, F, G, I — leaf nodes B C
• The root has level 0
• The depth of a node is its level in D E F G H
the tree

• The height of the tree is the

I
maximum level of any node

• The degree of a node is the number

of children
2
 Example Tree MD/CEO

Director- Director- Director- Director

Operations Marketing Accounts & Human
Finance Resources

GM- GM- GM- GM- GM- GM- GM-

Manufacturing Quality Sales CRM Accounts Finance HR

Chief Plant Chief Quality Chief Chief

Manager Manager Manager — Manager —
Sales CRM

Manager, Manger, Managers, Manager CRM,

Workers Workers Sales Executives CRM Executives
3
Example Tree

4
Example Tree

5
Binary Tree
• An ordered tree is one in which the children of each node are ordered.
• Binary tree is an ordered tree with all nodes having at most 2 children
A

B C

D F H

6
Recursive Definition of Binary Tree
• A binary tree is either a
• leaf or
A
• An internal node (the
root) and one/two
B C
binary trees (left subtree
and/or right subtree).
D F H