Last

class we discussed about hashing. We saw few collision resolution techniques,

chaining,
double hashing linear programming and you also did a little bit of analysis of
these collision resolution techniques. Today we are going to talk about trees. We
are also going to look at binary trees and some data structures for trees.
What is a tree? Many of you might have come across a tree before, except this
tree is
going to be different from one that you have seen before. The root will be at the
top.
In most of the trees around, you do not see the root. The root is going to be at
the top of the tree. In the tree given in the slide above
A is the root. There is a notion of parent and children, the node B is the parent of
node D and E. By the same argument A is the parent of B and C, C is parent of F,
G and
H. A is a parent of B which in turn parent of D and E, so A is ancestor of D and E.
A
is also an ancestor of F, G and H. A is also an ancestor of I. A is a grandparent,
sometimes
we use the term grandparent. A is a grandparent of D, E, F, G and H. Hope you
understand the difference between ancestor and grandparent. D and E are
descendents
of A, in fact B, C, D, E, F, G, H, I are all descendents of A.
C and B are siblings because they have the same parent. B is a sibling of C and C
is
a sibling of B. G and E are not siblings but F, G and H are siblings. D and E are
the children
of node B. A is a parent of B, B is a parent of D and
E, D and E are children of B, B and C are children of A and all of these are
descendents
of A. I have 3three ancestors H, C and A. H is the parent, C is the grandparent
and A is the
great grand parent but we do not use that term we just call it as an ancestor. The
terms
we defined till now were more in the nature of a family tree and then we will
come to real trees. D, E, F, G, I are called the leaves of the tree. A is the root, if
you just turn
it upside down then the extremities should be the leaves.
What is the leaf? The generic term for A, B...I are also called nodes of a tree. A
leaf
is a node which has no children. If a node has no children then it is a leaf. H is not
a leaf since H has a child but I, F, G, E and D are nodes which do not have any
children
and so they are called the leaves. A, B, C and H are called internal nodes, a node
which is not a leaf is called an internal
node. We associate a notion of level with each node, the root is at level 0. The
children
of the root are at level 1. The children of those nodes which are at level 1 are at
level
2. D, E, F, G and H all are at level 2. It is not that H is at level 2, all of these
nodes are at level 2. I is at level 3.
Sometimes we also use the term depth, in which depth and level are the same
thing. The level
0, level 1, level 2 are also called as depth 0, depth 1 and depth 2. The height of
the
tree is the maximum level of any node in the tree.
What is the maximum level of any node in the tree? The height of this tree is 3.
The degree of the node is the number of children it has. B has a degree 2, C has
a degree 3
and H has a degree 1. The leaves have degree 0 because they do not have any
children. Basic
terminologies are quite intuitive. What are trees used for?
They can represent the hierarchy in an organization. For instance there is a
company let us call
electronics R Us which has some divisions. RND is 1 division, purchasing is
another and
manufacturing is division. Domestic and international are the sub-division for the
sales. You could
represent the organizational structure through a tree. You could also use a tree to
represent
the table of contents in a book. Let us say a book called student guide which has
chapters on overview, grading, environment,
programming and support code. The chapter grading has some sections called
exams, homework
and programs. They could also have some sub-sections and that would build up
the tree.
Your file system in which if you use the Unix environment or the Windows
environment is
also organized as a tree. The one in the level 0 is the root directory and in the
1stlevel
are those 2 sub-directories. Then with in the sub-directory I have some other sub-
directories
and with in that I have homework, assignment and so on. Your file system is also
organized
like a tree. Today in our class we are going to see about definitions and then we
start
using those definitions in our later classes.
An ordered tree is one in which the children of the each node are ordered. That
means there
is a notion in which we would like to put the left child in the 1stlevel to the right
side. Suppose if you want to draw a family tree, you may want to draw the eldest
child
to the left and the younger child as you move from left to right.
There is a notion of order there and some time you want to reflect that order in
your tree. But there would be no notion of order in the following example.
The node which is in the level 0 is a directory and in the 1stlevel there are two
sub-directories. Whether I place the left node to the right or the right node to the
left, it does not
really make any difference as far as the picture is concerned. Also it does not
convey any additional information but sometimes you might have the notion of
order between the children.
Such a tree is called an ordered tree.
A binary tree is an ordered tree in which there is a notion of left child and a right
child. Actually it is an ordered tree in which every node has at most 2 children.
The diagram
given in the slide below is an example of a binary tree. The root node has 2
children and the child node on the left has only 1 child and the
following child on the right has only 1 child. The node which does not have child
node are
said to be leaves. We have 5 leaves which have no children. These nodes in the
1stlevel are ordered and there is a notion of left and right nodes. If I were to
change the tree
that is if I were to draw the left nodes on the right and right nodes on the left
then I get a different binary tree. That would still be a binary tree but it would be
different
from
this binary tree.
All of this is dependent upon the application you have. This is just a way of
representing information. Sometimes the order has meaning to it, sometime it
has no meaning to it. When
it has some meaning to it then you would rather use an ordered binary tree and
when you change
the order, then you are representing something different. We will see more
example of this
and things would become clear. I can also define a binary tree in a recursive form
as follows. A binary tree is just a single
node or a leaf or it is an internal node which is the root to which I have attached
2 binary
trees. In the following slide the nodes which are marked on the left side are
called left subtree and the nodes marked on the right are called right subtree. I
can construct
any binary tree in this manner. I take a node and attach a left subtree and a right
subtree. I get a left subtree and
right subtree through recursive in which it is obtained by taking a node and
attaching it to the left and right subtrees.
I have said and/or which means this left subtree might be null that is I might not
attached
anything or I might not attached anything to the right or I might have attached
both
the subtrees. Remember we have introduced other terms, left subtree and right
subtree. The node to the left side of the root node is called the left subtree and
the node on
the right side is called right subtree.
What is the left subtree of the node which is in the 1stlevel? In the 2ndlevel, the
node at the extreme left is the left subtree of the node.
One example of a binary tree is the arithmetic expressions. I have an arithmetic
expression
which looks like the one given in the slide below. I can represent this as a binary
tree.
Let us look at a parenthezisation of this expression. Suppose I have
parenthesized in
the manner like, which is given in the last line of the slide. We have (4+6), the
numbers here will be the leaves of my binary tree and the internal node would
correspond to
the operations. In fact this is also one way to evaluate this expression. You would
take
4+6 and you would sum that. You would draw a tree which has 1 internal node
and its two
children are 4 and 6. The internal node would have plus operator in it.
Whatever is the resulting value we are adding that to 1. I draw a tree whose root
is a plus
operator and one child is 1 and the other child is the subtree that is obtained
from
this operation and I could build this tree. This is just another way of representing
arithmetic
expression.
Decision tree is another example of binary tree. The example given in the slide
below is taken from the book. Star bucks, Café paragon and most of it would not
make much
sense, may be we would not come across them. What is the decision tree?
Each node in the decision tree corresponds to some decision that you want to
make. You
come to root node and ask whether you want a fast meal. The answer is yes then
you come
to the left node and whether you want coffee or not. The answer is yes then you
go to star bucks. If the answer is no you may go to some other place and so on.
Thus decision trees
are another example of binary trees. Why because typically it is yes and no. You
would follow
the decision tree to get into a particular node.
This was just more of terminology and examples. Let us see more concrete stuff.
Let us define
a complete binary tree. We are still at binary trees, as you can see every node in
this tree
has less than or equal to 2 children or at most 2 children. But I will call such a
tree
as a complete binary tree. We call a tree as a complete binary tree if at the level
there are nodes. In some sense it is full and when every node has 2 children
it does not give you a complete binary tree. I will show you why it cannot be a
complete binary tree.
Let us look at the slide below and check whether every internal node have 2
children in this
tree. Every node has a 2 children then that tree should also have leaves. It
cannot be the case in which every node has 2 children, in some case there are
no children. Just with
the requirement that every node has 2 children, every node other than the leaf
that means every internal node has 2 children does not implies it as a complete
binary tree. This
is a counter example to that in which every internal node has 2 children. This is
not
a complete binary tree.
The following is an example of a complete binary tree. We want to say that at
level i there are node. The root node is at level 0 that is 1 node, at level 1 there
are 2 nodes,
at level 2 there are 4 and at level 3 there are 8. If h is the height of the tree, in
the following example what is the height of the tree?
We call height as the maximum level number so we should not count this as 4.
Thus the
height of the tree is 3. If h is the height of the tree that means all the leaves are
at level h then by the definition of the binary tree
we have said that the level i has nodes that means there are leaves. The number
of leaves
in a complete binary tree of height h is just .
What is the number of internal nodes? At level 0 we have 1 node, at level 1 we
have
2 nodes and so on. Thus the sum is given as 1+2+ , because at level h all the
nodes are
leaf nodes. Thus the sum is ,
this is the number of internal nodes and the number of leaves is . The number of
internal nodes is the number of leaves-1. This is for a complete binary tree.
What is the total number of nodes in this tree? It is which is the number of leaves
+ which is the number of internal nodes, hence it
becomes . Let us call this number as n.
If I have a complete binary tree of n nodes, what is the height of this tree?
Let us go one step at a time. What is the number of leaves in this tree?
If the number of nodes is n and the number of leaves was which equals , just
from this
( =n) expression. The number of leaves in a complete binary tree on n nodes is .
If
I have a complete binary tree on n nodes, half of the nodes are leaves and the
remaining
half are of internal nodes. Similarly I can say that if I have a tree on n nodes,
then
the height of the tree is (no of leaves). I can evaluate h from ( =n), h will be log
( ) and so it is the log (no of leaves). Else we can go directly from , where the
number
of leaves is and so h is log (no of leaves).
You are just doing some simple counting here. If I give you a complete binary
tree of height
h then you should be able to say about the number of leaves and the number of
internal nodes it has. When I give you a complete binary tree on n nodes, you
should be able to say
the height and so on. If you have a tree on n nodes then the height of the tree is
log
( ). The other thing your have to keep in mind is that in such a tree the number of
leaves is very large. It is roughly half the total number of nodes. It is very leafy
kind of
a tree. So far we have seen a complete binary tree but a binary tree is any tree
in which every
node has atmost 2 children. To get any binary tree, you can start with a suitably
large
complete binary tree and just cut it off. For instance if I were to cut off some
pieces then I would get a binary tree as shown in
the slide below. I can always do it, no matter about the tree I need. Take the
binary tree
on the right side as height 3 then I would start with the complete binary tree of
height 3 which is on the left. Just cut off some pieces on the left side of the tree
to get
the tree which is on the right side. The picture given in the slide below is the
proof.
Let us use this fact that you can obtain any binary tree by just pruning of a
complete binary tree. Take a complete binary tree, cut off some branches then
you will get a
binary tree. If I have a binary tree of height h then in a complete binary tree at
level
i there were atmost nodes. In a binary tree at level i there will be atmost nodes,
there
cannot be more than nodes because the binary tree is obtained from a complete
binary tree by pruning. This is an important fact, atmost nodes at level i implies
that the total number of nodes
in your binary tree of height h is atmost 1+2+ nodes. The last level is h, at level
0 there will be 1 node, at level 1 there is atmost 2 nodes, at level 2 there are
atmost
4 nodes and so on. This is the maximum number of nodes that binary tree can
have.
Let us rewrite this. Suppose I told you that a tree has n nodes. Then n is less than
or
equal to this ( ) quantity, n <= which means that the height of the tree is just
rearranged
and it is h >= . If I give you a binary tree with n nodes in it, its height is atleast
and there is a particular binary tree which achieves this equality and that is a
complete binary tree.
Think of a complete binary tree as a tree which acquires the smallest height. If I
create
a binary tree with the certain number of nodes, the one which has the shortest
height will
be a complete binary tree. Because there we are packing all the nodes as close
to the root as possible by filling up all the levels to the maximum. That is the
minimum height
of the binary tree. I give you a binary tree on n nodes, its minimum possible
height is
. What is the maximum height that a binary tree on n nodes can have?
A binary tree on n nodes has height atmost n-1. This is obtained when every
node has
exactly 1 child and the picture is given in the slide below. This would be a zigzag
in
any manner and the height is 8 since there are 9 nodes in it.
In a binary tree on n nodes the minimum height is log (n) that is log ( ), but we
say it
as log (n) and the maximum height is n-1. That is the mistake many people
make. They
always assume that binary tree means height is log (n). But it is not the case, it
could be anywhere between log n and n.
How many leaves does the binary tree have? What is the minimum and the
maximum number
of leaves it can have? Let us figure it out. We will prove that the number of
leaves in
a tree is <=+ no of internal nodes. This is the useful inequality, in any binary
tree
the number of leaves is <+ the number of internal nodes or atmost the number
of leaves in a
tree can be 1 more than the number of internal nodes. How will you prove this?
We will prove it by induction on the number
of internal nodes. In a base case consider a tree with 1 node. If a tree has only 1
node how many internal
nodes does it have? It is 0, because that 1 node does not have any child so that
is the leaf. Base case is when the number of internal nodes is 0, in
which case the right hand side is 1 that is the number of leaves is 1 so the
inequality is satisfied.
We will assume that the statement is true for all trees with less than or equal to
k-1
internal nodes. This should be read as, the statement is true for trees with
atmost k-1
internal nodes not just k-1 but anything even for less this statement is true.
We will prove it for a tree with k internal nodes. Suppose I have a tree with k
internal
nodes, let us say on the left subtree I have internal node. Then how many
internal nodes
do I have on the right subtree? It is exactly k- -1 and not atmost because
all the internal nodes are either in the left subtree or in the right subtree or it is
the root node. The minus one is because of the root node. This is the number of
internal
nodes in the right subtree. Let us apply the induction hypothesis. is less than or
equal to k-1 and the quantity
(k- -1) is also less than or equal to k-1. We can use the induction hypothesis. In
the
left subtree which has internal nodes, the number of leaves is less than or equal
to +1. In the right subtree the number of leaves is less than or equal to k- -1+1
which is
k- . The total number of leaves is just the sum of these two (( +1) + (k- )), all the
leaves are either in the left subtree or in the right subtree. The total number of
leaves is just the sum which is k+1 that is we wanted to prove.
Since we started a tree with a k internal node, you have to show that the number
of leaves is less than 1+k. This is a simple proof which shows the number of
leaves is
atmost 1+ the number of internal nodes. There was a tree in which we saw the
number
of leaves is equal to the number of internal nodes +1. It was in a complete binary
tree.
What was the number of leaves in a complete binary tree? The number of leaves
was , if h was the height of the tree and the number of internal nodes
was 1+2+ nodes. There was exactly a difference of 1 between the number of
leaves and the
number of internal nodes. The complete binary tree once again achieves the
equality.
For any other tree the number of leaves will only be less than or equal to this
sum.
How small it can be? Let us look at that. For a binary tree on n nodes, the
number of leaves + the number
of internal nodes is n. Because every node is either a leaf or an internal node.
Also
we just saw that the number of leaves is less than or equal to the number of
internal nodes +1. I will just rearrange, this implies that the number of leaves is .
I have just rearranged,
as the number of internal nodes is greater than or equal to the number of leaves
-1.
I replace that and get the number of leaves as for any binary tree.
The another thing to keep in mind is for any binary tree the number of leaves will
never be more than half the number of nodes in the tree. Again this equality ( )
was achieved
for our complete binary tree, which is the most leafy tree. All others trees are dry
and the minimum number of leaves that tree might have is just 1. The example
for that is the same example that I have showed you before. The tree on 9 nodes
has only 1leaf in it. Let us look at an abstract data type for trees. You would have
the generic
methods which you seen for all the abstract data types till now. The following are
the
generic container methods, size () which tell us about how many nodes are there
in the tree, isEmpty () tells whether the tree is empty or not and the method
elements () which list
out all the elements of the tree.
You can have position based container methods, it is as the kind we saw for the
list or sequence
data types. The swapElements (p, q) in which I have specified 2 positions p and
q. Think
of the positions as references in to the tree except that using the position data
type I
am not able to access anything else but the elements sitting at that position.
The method positions () will specify all the positions in the tree. It will give you all
the positions in the tree as a sequence. The positions method has no
parameters, when you
invoke it on a certain tree it will just give you a sequence of all the positions in
the
tree, references to all the nodes in the tree. Once you access a particular position
then
using the element method on that positions you can access the element in that
node.
The swapElements (p, q) given 2 positions p and q, you are swapping the
elements at
these 2 positions. replaceElement (p, e) which means that given a position p you
are replacing
the element at that position with e. In query methods given a particular position
isRoot
(P) is this the root of the tree. Given a particular position isInternal (p) is this
is an internal node, given a particular position isExternal (p) is this external or
leaf. Sometimes
we use external or sometimes we use leaf, does this position correspond to leaf.
In accessor method when I call root it will return a position of the root, an object
of
type position. isRoot (p) is determined as given a position is it a root and root ()
returns
the position of the root. Hope you understand the difference between the both.
The position
of the root means it is not a reference to that particular node but it is a reference
of type position so that you cannot access anything except the element. This was
the
same as the type casting which we did earlier. The method parent (p), given a
particular position returns the parent node. The children
(p), given a particular position returns the children of this node.
If it were children, there could be of many children for a certain node.
How it will return the various children? It will return as a sequence, it will return
a sequence of object type sequence which will contain the position of all the
children.
Position has an element method which will let you to access the data. The update
methods
are typically application specific and this would be the generic method for a tree.
Binary tree should really be treated as a sub-class, as a derived class from a tree.
All we need to do is to continue to have the same method as we described for
the tree but we will have some additional methods. There would be a notion of a
left child given a
position give me the left child, give me the right child or give me a sibling.
We will come to the update methods when we see the example of it.
What is the node structure in a binary tree? What are the kinds of data that you
would
be keeping in an object corresponding to a node of the binary tree?
You would have the data, you would have a reference to the left child and a
reference
to the right child and you would have reference to the parent typically. You would
also have
a reference to key or data associated with this node, any element that is sitting
in
this node you would have reference to it and these were all sitting together. The
reference
to the node which is at the center will not be stored in the node and that does
not make
any sense. For instance if I access to this node, suppose this was the root node
and I use the root
method to get a position to this node then using that position I can now access
the left
child by invoking the left child method and in this manner I can get the position
of any
node. Once I have the position of a node I can then invoke element method and
any method
to get the data associated with that. A node in this case would definitely
implement the
position interface. In the slide given below, this is what the binary tree would
look like if you look at
the links and so on. The parent link would be null for a root node because it has
no parent. Then it would have
left child in which the left child would be referring to the node on the left and the
right child would be referring to the node which is on the right and so on. In the
above
diagram the extreme right node does not have any right child, its right child
member would
be referring to null.
That was for a binary tree. How do we take care of arbitrary trees?
Let us say unbounded trees. The root node has 3 children and its child which is
on the
left also has 3 children. Are we going to have 3 different data members to refer
to 3 children? That is not clear
about how to do it, because then if it has 4 children then how you would create
space
for another member. The way to do it is that you have a reference to 1of the
child only
and then all the children are in a linked list. Each child will have a reference to
the parent, so all of these children would be pointing to the parent node. But the
parent node would be pointing to only 1of them. Which would be the head of the
linked list
in the 2nd level? From the node which is at the 1st level, if
I want to refer to the children, I can just come here essentially return all the
elements of this linked list.
How do I know that I have reached the last element of the linked list when the
next is empty? The 1stfield of the node is empty because
it does not have children. Every node still has only 3 data members, parent or 3
references
1 for the parent, 1 for left most child and 1 for the right sibling. The left node on
the level 1 would refer to left most child and not to all the children because that
we
do not know how many are there. It will have 1 more to refer to the right sibling
because
for the left node at the level 2 should refer to the right sibling.
You can do with only 3 references. The node in the level 2 has only 1 child and it
is
just pointing to that 1 child, there is no sense of left and right here. This is not
a binary tree, left and right makes sense only in a binary tree.
Actually I should not have written left child, I should have written 1st child. That
is any
1 child then it just point to that 1 child and that let you to access its siblings
through
a linked list. From the 1st child you will go to the next child and to the next child
and so on. You can step through all the various children throughout linked list.
With that
we will end our discussion on binary trees today. In the next class we are going to
look
at reversals
of trees.
English - Lec 06