CST 306 - Algorithm Analysis and Design(S6 CSE)

Module IV
 Dynamic Programming, Back Tracking and Branch & Bound
o Dynamic Programming Control Abstraction
 The Optimality Principle
 Matrix Chain Multiplication-Analysis
 All Pairs Shortest Path Algorithm - Floyd-Warshall Algorithm-Analysis
o The Control Abstraction of Back Tracking – The N Queen’s Problem
o Branch and Bound Algorithm for Travelling Salesman Problem

 Dynamic Programming
o Dynamic programming is an algorithm design method that can be used when the solution to a
problem can be viewed as the result of a sequence of decisions.
o Dynamic Programming is mainly an optimization over plain recursion.
o Wherever we see a recursive solution that has repeated calls for same inputs, we can optimize
it using Dynamic Programming.
o The idea is to simply store the results of subproblems, so that we do not have to re-compute
them when needed later.
o This simple optimization reduces time complexities from exponential to polynomial.
o For example, if we write simple recursive solution for Fibonacci Numbers, we get exponential
time complexity and if we optimize it by storing solutions of subproblems, time complexity
reduces to linear.
o In dynamic programming an optimal sequence of decisions is obtained by making explicit
appeal to the principle of optimality.

o The Optimality Principle

 Definition: The principle of optimality states that an optimal sequence of decisions has
the property that whatever the initial state and decision are, the remaining decisions must
constitute an optimal decision sequence with regard to the state resulting from the first
decision.
 A problem is said to satisfy the Principle of Optimality if the subsolutions of an optimal
solution of the problem are themesleves optimal solutions for their subproblems.
 Examples:
 The shortest path problem satisfies the Principle of Optimality.
 This is because if a,x1,x2,...,xn,b is a shortest path from node a to node b in a graph,
then the portion of xi to xj on that path is a shortest path from xi to xj.

o Characteristics of Dynamic Programming

1. Overlapping Subproblems
 Subproblems are smaller versions of the original problem. Any problem has
overlapping sub-problems if finding its solution involves solving the same
subproblem multiple times.
 Dynamic Programming also combines solutions to sub-problems. It is mainly used
where the solution of one sub-problem is needed repeatedly. The computed solutions
are stored in a table, so that these don‟t have to be re-computed. Hence, this technique
is needed where overlapping sub-problem exists.
 For example, Binary Search does not have overlapping sub-problem. Whereas
recursive program of Fibonacci numbers have many overlapping sub-problems.

1 CS KTU Lectures
 CST 306 - Algorithm Analysis and Design(S6 CSE)

2. Optimal Substructure
 A given problem has Optimal Substructure Property, if the optimal solution of the
given problem can be obtained using optimal solutions of its sub-problems.
 For example, the Shortest Path problem has the following optimal substructure
property: If a node x lies in the shortest path from a source node u to destination
node v, then the shortest path from u to v is the combination of the shortest path
from u to x, and the shortest path from x to v.

o Steps of Dynamic Programming

 Dynamic programming design involves 4 major steps:
1. Characterize the structure of an optimal solution.
2. Recursively define the value of an optimal solution.
3. Compute the value of an optimal solution, typically in a bottom-up fashion.
4. Construct an optimal solution from computed information.

o Optimal matrix multiplication

 Suppose we wish to compute the product of 4 matrices A1 x A2 x A3 x A4
 The different parenthesizations are
( A1(A2( A3 A4)))
( A1((A2 A3) A4))
( (A1A2)( A3 A4))
(( A1(A2 A3))A4)
((( A1A2) A3)A4)

 We can multiply two matrices A and B if and only if they are compatible: The number of
columns of A must be equal to the number of rows of B.
 If A is a pXq matrix and B is a qXr matrix, the resulting matrix C is a pX r matrix.
 The time to compute C is pqr.
 We shall express costs in terms of the number of scalar multiplications
 Example:
 Consider 3 matrices A1, A2 and A3. Its dimensions are 10X100, 100X5, 5X50
respectively.
 Number of scalar multiplications for
o ( (A1A2) A3) is 7500
o (A1 (A2A3)) is 75000
 Thus, computing the product according to the first parenthesization is 10 times faster.

 Matrix-Chain Multiplication Problem: Given a chain (A1, A2, . . . An ) of n matrices,

where for i = 1, 2, . . . n, matrix Ai has dimension pi-1 X pi , fully parenthesize the product
A1 A2. . . An in a way that minimizes the number of scalar multiplications
 In the matrix-chain multiplication problem, we are not actually multiplying matrices.
 Our goal is only to determine an order for multiplying matrices that has the lowest cost

2 CS KTU Lectures
 CST 306 - Algorithm Analysis and Design(S6 CSE)
 Matrix Chain Multiplication : Dynamic Programming Method
 Step 1: The structure of an optimal parenthesization
o Ai. .j denote the matrix that results from evaluating the product AiAi+1 . . . Aj where
i<= j
o If i< j, we must split the problem into two subproblems (Ai Ai+1 . . . Ak and
Ak+1Ai+1 . . . Aj ), for some integer k in the range i<= k < j.
o That is, for some value of k, we first compute the matrices Ai. .k and Ak+1. .j. Then
multiply them together to produce the final product Ai. .j .
o Total cost = Cost of computing the matrix Ai. .k+ Cost of computing Ak+1. .j+ Cost
of multiplying them together.

 Step 2: A recursive solution

o We can define the cost of an optimal solution recursively in terms of the optimal
solutions to subproblems.
o Let m[i, j] be the minimum number of scalar multiplications needed to compute
the matrix Ai. .j
o For the full problem, the lowest cost way to compute A1. .n would thus be m[1, n]
o Ai. .i = Ai so m[i,i] = 0 for i=1,2,. .. . n

o s[i,j] be a value of k at which we split the product AiAi+1 . . . Aj in an optimal

parenthesization.
o This will take exponential time

 Step 3: Computing the optimal costs

o Compute the optimal cost by using a tabular, bottom-up approach
Algorithm Matrix_Chain_Order(p)
{ n = p.length – 1
Let m[1..n, 1..n] and s[1..n-1 , 2..n] be new tables
for i=1 to n do
m[i,i] = 0
for l=2 to n do
{ for i=1 to n-l+1 do
{ j=i+l-1
m[i,j] = α
for k=i to j-1 do
{ q = m[i,k] + m[k+1,j] + pi-1 pk pj
if q<m[i,j] then
{ m[i,j] = q
s[i,j] = k
}
}

}
}
return m[][] and s[][]
}

3 CS KTU Lectures
 CST 306 - Algorithm Analysis and Design(S6 CSE)
o Matrix Ai has dimensions pi-1 X pi for i =1,2, . . . n.
o Input to this algorithm is a sequence p = ( p0, p1, . . . . pn ), where p.length = n + 1.
o The procedure uses 2 auxiliary tables
o m[1 .. n , 1...n] for storing the cost of matrix multiplication
o s[1..n-1 , 2 ...n] records which index of k achieved the optimal cost in computing
m[i,j] .

 Step 4: Constructing an optimal solution

Algorithm Print_Optimal_Parens(s,i,j)
{
if i==j then
print “A”i
else
print “(“
print_Optimal_Parens(s,i,s[i,j])
print_Optimal_Parens(s,s[i,j]+1,j)
print “)”
}
o Initial call is PRINT-OPTIMAL-PARENS(s,1,n)

 Time Complexity
 We are generating n(n-1)/2 number of elements in matrix m[].
 To calculate each element it will take atmost n time.
 So the time complexity = O( n.n(n-1)/2) = O(n3)

o Examples
1. Using Dynamic Programming, find the fully parenthesized matrix product for multiplying
the chain of matrices< A1 A2 A3 A4 A5 A6 > whose dimensions are <30X35>,
<35X15>, <15X5>, <5X10>, <10X20> and <20X25> respectively
2. Given a chain of 4 matrices <A1,A2,A3,A4> with dimensions
<5X4>,<4X6>,<6X2>,<2X7> respectively. Using Dynamic programming find the
minimum number of scalar multiplications needed and also write the optimal
multiplication order.
3. Find an optimal paranthesization of a matrix-chain product whose sequence of
dimensions is 4x10,10x3,3x12,12x20,20x7

o All pairs shortest path problem

 Find the shortest distances between every pair of vertices in a given weighted directed
Graph
 The Floyd Warshall Algorithm is for solving the All Pairs Shortest Path problem.
Negative edge weights are also allowed.
 As a result of this algorithm, it will generate a matrix, which will represent the minimum
distance from any node to all other nodes in the graph.

4 CS KTU Lectures
 CST 306 - Algorithm Analysis and Design(S6 CSE)
 Floyd Warshall Algorithm
Inputs are the adjacency matrix of the given graph and total number of vertices

Algorithm FloydWarshall(cost[][], n)
{
for i=1 to n do
for j=1 to n do
D[i, j] = cost[i, j]
for k := 1 to n do
for i := 1 to n do
for j := 1 to n do
D[i, j] = min{D[i, j] , D[i, k] + D[k, j] }
Return D
}

 Time Complexity
 Floyd Warshall Algorithm consists of three loops over all the nodes. Each loop has
constant complexities.
 Hence, the time complexity of Floyd Warshall algorithm = O(n3), where n is the
number of nodes in the given graph.
 Example
 Consider the following directed weighted graph. Using Floyd Warshall Algorithm, find
the shortest path distance between every pair of vertices

 Solution
o Remove all self loops and parallel edges(keeping the lowest weight edge) of the
given graph
o Write the adjacency matrix

o Find the matrix D1

 Keep the 1st row, 1st column and diagonal elements of D0 as such
 D1(2,3) = min{ D0(2,3), D0(2,1) + D0(1,3) } = min{α, 6+(-4)) = 2
 D1(2,4) = min{ D0(2,4), D0(2,1) + D0(1,4) } = min{2, 6+α) = 2
 D1(3,2) = min{ D0(3,2), D0(3,1) + D0(1,2) } = min{5, α +9) = 5
 D1(3,4) = min{ D0(3,4), D0(3,1) + D0(1,4) } = min{α, α + α) = α
 D1(4,2) = min{ D0(4,2), D0(4,1) + D0(1,2) } = min{α, α +9) = α
 D1(4,3) = min{ D0(4,3), D0(4,1) + D0(1,3) } = min{1, α +(-4)) = 1

5 CS KTU Lectures
 CST 306 - Algorithm Analysis and Design(S6 CSE)

o Find the matrix D2

 Keep the 2nd row, 2nd column and diagonal elements of D1 as such
 D2(1,3) = min{ D1(1,3), D1(1,2) + D1(2,3) } = min{-4, 9+2) = -4
 D2(1,4) = min{ D1(1,4), D1(1,2) + D1(2,4) } = min{α, 9+2) = 11
 D2(3,1) = min{ D1(3,1), D1(3,2) + D1(2,1) } = min{α, 5+6) = 11
 D2(3,4) = min{ D1(3,4), D1(3,2) + D1(2,4) } = min{α, 5+2) = 7
 D2(4,1) = min{ D1(4,1), D1(4,2) + D1(2,1) } = min{α, α+6) = α
 D2(4,3) = min{ D1(4,3), D1(4,2) + D1(2,3) } = min{1, α+2) = 1

o Find the matrix D3

 Keep the 3rd row, 3rd column and diagonal elements of D2 as such
 D3(1,2) = min{ D2(1,2), D2(1,3) + D2(3,2) } = min{9, -4+5) = 1
 D3(1,4) = min{ D2(1,4), D2(1,3) + D2(3,4) } = min{11, -4+7) = 3
 D3(2,1) = min{ D2(2,1), D2(2,3) + D2(3,1) } = min{6, 2+11) = 6
 D3(2,4) = min{ D2(2,4), D2(2,3) + D2(3,4) } = min{2, 2+7) = 2
 D3(4,1) = min{ D2(4,1), D2(4,3) + D2(3,1) } = min{α, 1+11) = 12
 D3(4,2) = min{ D2(4,2), D2(4,3) + D2(3,2) } = min{α, 1+5) = 6

o Find the matrix D4

 Keep the 4th row, 4th column and diagonal elements of D3 as such
 D4(1,2) = min{ D3(1,2), D3(1,4) + D3(4,2) } = min{1, 3+6) = 1
 D4(1,3) = min{ D3(1,3), D3(1,4) + D3(4,3) } = min{-4, 3+1) = -4
 D4(2,1) = min{ D3(2,1), D3(2,4) + D3(4,1) } = min{6, 2+12) = 6
 D4(2,3) = min{ D3(2,3), D3(2,4) + D3(4,3) } = min{2, 2+1) =2
 D4(3,1) = min{ D3(3,1), D3(3,4) + D3(4,1) } = min{11, 7+12) = 11
 D4(3,2) = min{ D3(3,2), D3(3,4) + D3(4,2) } = min{5, 7+6) =5

o D4 represents the shortest distance between each pair of the given graph

6 CS KTU Lectures
 CST 306 - Algorithm Analysis and Design(S6 CSE)
 Comparison of Divide and Conquer and Dynamic Programming strategies
o Both techniques split their input into parts, find sub-solutions to the parts, and synthesize
larger solutions from smaller ones.
o Divide and Conquer splits its input at pre-specified deterministic points (e.g., most probably in
the middle)
o Dynamic Programming splits its input at every possible split points rather than at pre-specified
points. After trying all split points, it determines which split point is optimal.
o Divide & Conquer algorithm partition the problem into disjoint subproblems. Solve the
subproblems recursively and then combine their solution to solve the original problems.
o Dynamic Programming is used when the subproblems are not independent, e.g. when they
share the same subproblems. In this case, divide and conquer may do more work than
necessary, because it solves the same sub problem multiple times.
o Dynamic Programming solves each subproblems just once and stores the result in a table so
that it can be repeatedly retrieved if needed again.

 Greedy vs. Dynamic Programming

o Both techniques are optimization techniques, and both build solutions from a collection of
choices of individual elements.
o The greedy method computes its solution by making its choices in a serial forward fashion,
never looking back or revising previous choices.
o Dynamic programming computes its solution bottom up by synthesizing them from smaller
subsolutions, and by trying many possibilities and choices before it arrives at the optimal set
of choices.
o There is no a priori litmus test by which one can tell if the Greedy method will lead to an
optimal solution.
o By contrast, there is a litmus test for Dynamic Programming, called The Principle of
Optimality
o The greedy method only generated one decision sequence ever.
o In dynamic programming, many decision sequences may be generated. However, sequences
containing suboptimal subsequences cannot be optimal and so will not generated.

 Back Tracking
o Backtracking method expressed the solution as n-tuple (x1, x2.,…… xn), where xi‟s are chosen
from some finite set Si.
o The problem to be solved calls for finding one vector that maximizes (or minimizes or
satisfies) a criterion function P(x1, x2.,…… xn).
o Examples: Sorting the array of integers in a[l: n]
 The solution to the problem is expressed an n-tuple, where xi is the index of the ith
smallest element.
 The criterion function P is: a[xi] ≤ a[xi+1], for 1≤ i < n.
 The set Si ={1,2, . . . . .n}
 Different methods for solving this problem
 Brute Force approach
o Suppose mi is the size of set Si.
o The number of tuples (with size n) that are possible candidates for satisfying the
function P is: m = m1 x m2 x m3 . . . . . . x mn
o Brute Force approach evaluates each one with P, and save those which yield the
optimum.

7 CS KTU Lectures
 CST 306 - Algorithm Analysis and Design(S6 CSE)
 Backtracking algorithm
o It yields the same answer with far fewer than m trials.
o Its basic idea is to build up the solution vector one component at a time and to use
modified criterion functions (bounding functions) P i(x1, x2.,…… xi ) to test
whether the vector being formed has any chance of success.
o The major advantage of this method is that if it is realized that the partial vector
(x1, x2.,…… xi) can in no way lead to an optimal solution, then mi+1 x mi+2. . . . .
. x mn possible test vectors can be ignored entirely.

o Backtracking method require that all the solutions satisfy a complex set of constraints. These
constraints can be divided into two categories:

 Explicit Constraints
 Explicit constraints are rules that restrict each xi to take on values only from a given
set
 The explicit constraints depend on the particular instance I of the problem being
solved. All tuples that satisfy the explicit constraints define a possible solution space
for I.
 Example:

 Implicit Constraints
 These are rules that determine which of the tuples in the solution space of I satisfy the
criterion function (Bounding Function).

o N-Queens Problem
 n queens are to be placed on a n x n chessboard so that no two attack. That is, no two
queens are on the same row, column, or diagonal.
 Number the rows and columns of the chessboard 1through n.
 The queens can also be numbered 1through n.
 Since each queen must be on a different row, we can assume that queen i is to be placed
on row i.
 All solutions to the n-queens problem can therefore be represented as n-tuples(x1,
x2.,…… xn), where xi is the column on which queen i is placed.
 Explicit constraint: Si = {1,2,3, . . . . n }, 1≤ i ≤n
 The solution space contains |S1| x |S2| x. . . . . . . x |Sn| = nn tuples.
 Implicit constraints:
 No two xi‟s can be the same(i.e. all queens must be on different columns)
o The solution space contains |S1| x |S2| x . . . . . . . x |Sn| = n(n-1) . . . . .1 = n! tuples
o It reduces the solution space from nn to n!.
 No two queens can be on the same diagonal.
1 2 3 4
1 Q1
2 Q2
3 Q3
4 Q4

8 CS KTU Lectures
 CST 306 - Algorithm Analysis and Design(S6 CSE)
 Following is a tree organization (permutation tree/State Space Tree) for 4-queen
problem without considering the last implicit constraint.

 The edges are labeled by possible values of xi.

 Edges from level 1 to level 2 nodes specify the values for x1
 Edges from level i to level i +1 are labeled with the values of xi
 Thus, the leftmost sub-tree contains all solutions with x1 = 1; its leftmost sub-tree
contains all solutions with x1 = 1 and x2 = 2, and soon.
 The solution space is defined by all paths from the root node to a leaf node. There are
4!= 24 leaf nodes in the above tree.

o Permutation tree or State Space Tree

 Tree organization of solution space is called state space tree.
 Problem State: Each node in this tree defines a problem state.
 State Space of the problem: All paths from the root to other nodes define the State
Space of the problem.
 Solution States: Those problem states s for which the path from the root to s defines a
tuple in the solution space.
 Answer states: Those solution states s for which the path from the root to s defines a
tuple that satisfies all implicit constraints of the problem.
 Live Node: A node which has been generated and all of whose children have not yet been
generated is called a live node.
 E-node : The live node whose children are currently being generated is called the E-node
 Dead node: It is a generated node which is not to be expanded further or all of whose
children have been generated.
 Bounding functions are used to kill live nodes without generating all their children.
 Problem States can be generated by:
 Depth First generation of the problem states:
o As soon as a new child C of the current E-node R is generated, this child will
become the new E-node.
o Then R will become the E-node again when the sub-tree C has been fully
explored.

9 CS KTU Lectures
 CST 306 - Algorithm Analysis and Design(S6 CSE)

o Backtracking: Depth first node generation with bounding functions.

 Breadth First generation of the problem states:
o The -E-node remains the E-node until it is dead.
o Branch-and-bound methods: Breadth first node generation method with
bounding function is called Branch and Bound method. There are two alternatives
 Breadth First Generation Method: Each new node is placed into a queue.
When all the children of the current-E-node have been generated, the next
node at the front of the queue becomes the new E-node.

 D-search(depth search): Each new node is placed into a stack. When all the
children of the current-E-node have been generated, the next node at the top of
the stack becomes the new E-node.

 At the conclusion of the process at least one answer node is always generated or all
answer nodes are generated if the problem requires us to find all solutions.

o Backtracking works on 4-Queens Problem

 If (x1, x2.,…… xi) is the path to the current E-node, then all children nodes with parent-
child labeling xi+1 are such that (x1, x2.,…… xi+1) represents a chessboard configuration in
which no two queens are attacking.
 We start with the root node as the only live node. This becomes the E-node and the path is
(). We generate one child(node 2) and the path is now (1). This corresponds to placing
queen1on column1.
 Node2 becomes the E-node. Node3 is generated and immediately killed.
 The next node generated is node8 and the path becomes(1,3). Node8 becomes the E-node.
However, it gets killed as all its children represent board configurations that cannot lead
to an answer node.
 We back track to node2 and generate another child node13.The path is now (1,4). This
process continues until it will generate a proper arrangement.

10 CS KTU Lectures
 CST 306 - Algorithm Analysis and Design(S6 CSE)

State Space Tree of 4 Queens Problem

o Backtracking Control Abstraction
 (x1, x2.,…… xi) be a path from the root to a node in a state space tree.
 Generating Function T(x1, x2.,…… xi) be the set of all possible values for xi+1 such that
(x1, x2.,…… xi+1) is also a path to a problem state.
 T(x1, x2.,…… xn) = φ
 Bounding function Bi+1(x1, x2.,…… xi+1) is false for a path (x1, x2.,…… xi+1) from the
root node to a problem state, then the path cannot be extended to reach an answer node.
 Thus the candidates for position i+1of the solution vector (x1, x2.,…… xn) are those
values which are generated by T and satisfy Bi+1.
 The recursive version is initially invoked by Backtrack(1).
Backtracking Control Abstraction
Algorithm Backtrack(k)
{
for (each x[k] ϵ T(x[1], . . . .x[k-1])
{
if(Bk(x[1], x[2], . . . . . ., x[k]) != 0) then
{
if(x[1], x[2], . . . . . ., x[k] is a path to an answer node)
then write(x[1:k])
if(k<n) then Backtrack(k+1)
}
}
}

11 CS KTU Lectures
 CST 306 - Algorithm Analysis and Design(S6 CSE)
th
o All the possible elements for the k position of the tuple that satisfy Bk are generated one by
one, and adjoined to the current vector (x1, x2.,…… xk-1 ).
o Each time xk is attached, a check is made to determine whether a solution has been found.
Then the algorithm is recursively invoked.
o When the for loop is exited, no more values for xk exist and the current copy of Backtrack
ends. The last unresolved call now resumes.
o This algorithm causes all solutions to be printed. If only a single solution is desired, then a flag
can be added as a parameter to indicate the first occurrence of success.

 N-Queens Problem(Cond…)
o Consider an n x n chessboard and try to find all possible way to place n non-attacking
queens.
o (x1, x2.,…… xn ) be the solution vector. Queen i is placed in ith row and xith column. xi will
all be distinct since no two queens can be placed in the same column.
o There are 2 type of diagonals
 Positive Diagonal
 Diagonal from upper left to lower right
 Every element on the same diagonal has the same row-column value
 Suppose 2 queens are place at position (i,j) and (k,l), then i-j = k-l
 Negative Diagonal
 Diagonal from upper right to lower left
 Every element on the same diagonal has the same row+column value
 Suppose 2 queens are place at position (i,j) and (k,l), then i+j = k+l
 The 1st equation implies: i-k = j-l
 The 2nd equation implies: i-k = l - j
 Combining these two, we will get : | i-k | = | j-l |
Absolute value of column difference is equal to the absolute value of row difference.

Algorithm NQueens(k,n)
{
for i=1 to n do
{ if Place(k,i) then
{ x[k] = i
if(k==n) then write(x[1:n])
else NQueens(k+1, n)
}
}
}

Algorithm Place(k,i)
{
for j=1 to k-1 do
{
if( (x[j]==i) or ( Abs(j-k)==Abs(x[j]-i) ) then
Return false
}
Return true
}

12 CS KTU Lectures
 CST 306 - Algorithm Analysis and Design(S6 CSE)

o Place(k,i) returns true if the kth queen can be placed in column i.

 i should be distinct from all previous values x[1],x[2], . . . . . . ., x[k-1]
 And no 2 queens are to be on the same diagonal
 Time complexity of Place() is O(k)
o NQueen() is initially invoked by NQueen(1,n)
 Time complexity of NQueen() is O(n!)

o Examples
 Show the state space tree for 4 Queens problem. Show the steps in solving 4 Queens
problem using backtracking method to print all the solutions

 Branch and Bound Technique

o During state space tree generation E-node remains E-node until it is dead.
o Two strategies:
 Breadth First Search(BFS)
 It is called FIFO(First In First Out). Here the live nodes are placed in a
queue.
 Depth Search(D-Search)
 It is called LIFO(Last In First Out). Here the live nodes are placed in a
stack.
o Least Cost Search(LC Search)
 To improve the searching speed, we can use a ranking function ĉ(.) for live
nodes.
 ĉ(.) value of each live node is calculated. The next E-node is the live node with
least ĉ(.). Such a search strategy is called LC Search.
 BFS and D-Search are the special cases of LC-Search.
 LC-Search coupled with bounding function is called LC Branch and Bound
Search.
o LC-Search Control Abstraction
Algorithm LCSearch(t)
{ if t is an answer node then output t and return
E=t
Initialize the list of live nodes to be empty
repeat
{ for each child x of E do
{
if x is an answer node then output the path from x to t and return
Add(x)
x  parent = E
}
if there are no more live nodes then
{ Write “no answer node”
return
}
E = Least()
}until(false)
}

13 CS KTU Lectures
 CST 306 - Algorithm Analysis and Design(S6 CSE)

 Least() finds a live node with least ĉ. This node is deleted from the list of live
nodes and returned.
 Add(x) adds the new live node x to the list of live nodes.
 LCSearch outputs the path from the answer node to the root t.
 LCSearch terminates only when either an answer node is found or the entire
state space tree has been generated and searched.
 The control abstraction for LC, FIFO and LIFO are same. The only difference is
the implementation of the list of live nodes.
 FIFO Search scheme:
o The list of live nodes is implemented as queue.
o Least() and Add(x) being algorithms to delete an element from
and add an element to the queue.
 LIFO Search scheme:
o The list of live nodes is implemented as stack.
o Least() and Add(x) being algorithms to delete an element from
and add an element to the stack.
 LC-Search Scheme:
o Add(x) is an algorithm to add elements to the list. Least() returns
a live node with least ĉ(.) from the list.

o Branch and Bound Algorithm for Travelling Salesman Problem

 Given a set of cities and distance between every pair of cities, the problem is to find the
shortest possible tour that visits every city exactly once and returns to the starting point.
 Example: Apply branch and bound algorithm to solve TSP for the following graph,
assuming the start city as „a‟. Draw the state space tree.

 Solution
 The adjacency matrix is

 Perform row reduction, then column reduction

14 CS KTU Lectures
 CST 306 - Algorithm Analysis and Design(S6 CSE)
Total cost reduced = 2+2+1+1+0+0+0+0 = 6
The state space tree is

M1 is the matrix for node 1 is

 Generate the child node of node 1

 Find the matrix and cost of node 2

o Set row a and column b elements are α
o Set M1[b, a]= α
o The resultant matrix is

o Perform row reduction, then column reduction

Cost reduced = r = 5
M2 is the matrix for node 2
Cost of node 2 = Cost of node 1 + M1[a, b] + r = 6 + 0 + 5 = 11
 Find the matrix and cost of node 3
o Set row a and column c elements are α
o Set M1[c, a]= α
o The resultant matrix is

o Perform row reduction, then column reduction

Cost reduced = r = 2

15 CS KTU Lectures
 CST 306 - Algorithm Analysis and Design(S6 CSE)
M3 is the matrix for node 3
Cost of node 3 = Cost of node 1 + M1[a, c] + r = 6 + 3 + 2 = 11
 Find the matrix and cost of node 4
o Set row a and column d elements are α
o Set M1[d, a]= α
o The resultant matrix is

o Perform row reduction, then column reduction

Cost reduced = r = 6
M4 is the matrix for node 4
Cost of node 4 = Cost of node 1 + M1[a, d] + r = 6 + 5 + 6 = 17
o Now the state space tree is

Now the live nodes are 2, 3 and 4. Minimum cost is for node 2 and 3. Choose
one node(say node 2) as the next E-node.
o Generate the child node of node 2

 Find the matrix and cost of node 5

o Set row b and column c elements are α
o Set M2[c, a]= α
o The resultant matrix is

16 CS KTU Lectures
 CST 306 - Algorithm Analysis and Design(S6 CSE)
o Perform row reduction, then column reduction

Cost reduced = r = 2
M5 is the matrix for node 5
Cost of node 5 = Cost of node 2 + M1[b, c] + r = 11 + 5 + 2 = 18
 Find the matrix and cost of node 6
o Set row b and column d elements are α
o Set M2[d, a]= α
o The resultant matrix is

o Perform row reduction, then column reduction

Cost reduced = r = 0
M6 is the matrix for node 6
Cost of node 6 = Cost of node 2 + M1[b, d] + r = 11 + 0 + 0 = 11

o Now the state space tree is

Now the live nodes are 5, 6, 3 and 4. Choose one node which having minimum
cost(say node 6) as the next E-node.

17 CS KTU Lectures
 CST 306 - Algorithm Analysis and Design(S6 CSE)
o Generate the child node of node 6

 Find the matrix and cost of node 7

o Set row d and column c elements are α
o Set M6[c, a]= α
o The resultant matrix is

o Perform row reduction, then column reduction

Cost reduced = r = 0
M7 is the matrix for node 7
Cost of node 7 = Cost of node 6 + M6[d, c] + r = 11 + 0 + 0 = 11

o Now the state space tree is

18 CS KTU Lectures
 CST 306 - Algorithm Analysis and Design(S6 CSE)
Now the live nodes are 5, 7, 3 and 4. Choose one node which having minimum
cost(say node 7) as the next E-node. Node 7 having no child node.
Now we can say that node 1 to node 7 path is the Traveling salesperson path.

The TSP path = a-b-d-c-a

The TSP cost = 11

19 CS KTU Lectures