3.

1 Height-Balanced Trees
3.2 Weight-Balanced Trees

Chen Song
Height-Balanced Trees
 Height-Balanced Trees

Height – The maximum length of any path from the root

to a leaf.

Height-Balanced Tree –
In each interior node, the height of the right
subtree and left subtree differ by at most one.
 Height-Balanced Trees

Example:
 Height-Balanced Trees

 Theorem
A height-balanced tree of height h has at least
3+ 5 1+ 5 ℎ 3− 5 1− 5 ℎ
( )( ) −( )( )
2 5 2 2 5 2
leaves.
 Height-Balanced Trees

Tree T

h-2
h-1

Tree T has at least leaves:

leaves(h)=leaves(h-1)+leaves(h-2)
leaves(0)=1
leaves(1)=2
Characteristic equation:
xh=xh-1+xh-2 => x2-x-1=0
 Height-Balanced Trees

 Insert and Delete

After insert or delete:

Case 1: Tree is still balanced
Case 2: Tree is not balanced at node n
|n->left-height − n->right-height|=2
 Height-Balanced Trees

1. n->left->height = n->right->height+2 and

n->left->left->height=n->right->height+1
Right rotation on node n

2. n->left->height = n->right->height+2 and

n->left->left->height=n->right-height
Left rotation on node n->left, and follow right rotation
on node n
 Height-Balanced Trees

3. n-> right->height = n->left->height+2 and

n->right->right->height=n->left->height+1
Left rotation on node n->left

4. n->right->height = n->left->height+2 and

n->right->right->height=n->left-height
Right rotation on node n->right, and follow left rotation
on node n
 Height-Balanced Trees

 Height-Balanced Tree structure supports search, insert, and

delete in O(logn) time

Search => O(logn)

Insert => search + insert + rebalance => O(logn)
O(logn) O(1) O(logn)

Delete => search + delete + rebalance => O(logn)

O(logn) O(1) O(logn)
Weight-Balanced Trees
 Weight-Balanced Trees

Weight –
The number of leaves of a tree.

Weight-Balanced Tree –
The weight of the right and left subtree in each
node differ by at most one.
 Weight-Balanced Trees

 α-weight-balanced trees
For each subtree, the left and right sub-subtrees has
each at least a fraction of α of total weight of the subtree.
Tree T:
αWT≤WT1≤(1-α)WT
αWT≤WT2≤(1-α)WT
T2
T1
 Weight-Balanced Trees

 Theorem
An α-weight-balanced tree of height h≥2 has at least
1 ℎ
( ) leaves.
1−α
 Weight-Balanced Trees

 Rebalance
2 1
n is current node α∈[ , 1 − ]
7 2

Case 1:
n->left->weight ≥ α*n->weight and
n->right->weight ≥ α*n->weight
No rebalancing
 Weight-Balanced Trees

Case 2:
n->right->weight ≥ α*n->weight
If n->left->left->weight > (α+ε)n->weight,
do right rotation on node n
Else left rotation on n->left, followed right rotation on
node n.
2 1
ε≤ α -2α+
2
 Weight-Balanced Trees

Case 3:
n->left->weight ≥ α*n->weight
If n->right->right->weight > (α+ε)n->weight,
do left rotation on node n
Else right rotation on n->right, followed left rotation
on node n.
 Weight-Balanced Trees

 Theorem
The weight-balanced tree structure supports search, insert,
and delete in O(logn) time.

Search => O(logn)

Insert => search + insert + rebalance => O(logn)
O(logn) O(1) O(logn)

Delete => search + delete + rebalance => O(logn)

O(logn) O(1) O(logn)
END