Artificial Intelligence

Outline
 Last Class
 Properties of Search Algorithms
 Uninformed Search Algorithms
 Breadth-first search
 Depth-first search
 Depth-limited search
 Uniform-cost search
 Iterative deepening search

 Today
 Informed Search Algorithms
 Heuristics function
 Best first search algorithm
 Greedy Best First search
 A* search
 Uninformed Search Algorithms
 Use only the information available in the problem definition
 Carried out without any additional information that is
already provided in the problem statement
 How to traverse the tree, how to identify leaf and goal nodes
 Solution cost is not taken into account

 Breadth-first search
 Depth-first search
 Depth-limited search
 Uniform-cost search
 Iterative deepening search
 Informed Search Algorithms
 Informed search algorithm have information on the
goal state such as how far we are from the goal, path
cost, how to reach to goal node, etc.
 This information helps agents to explore less to the
search space and find the goal node, more efficiently.
 Searches the most promising branches of the state-
space first

 This information is obtained by something called

a heuristic, so it is also called Heuristic search.
 Heuristics function
 In an informed search, a heuristic is a function that
estimates how close a state is to the goal state.
 Informs about which nodes in the state-space to
expand next that lead quickly to the goal state.

 Heuristics function is represented by h(n)

 For example – Manhattan distance, Euclidean

distance, etc. (Lesser the distance, closer the goal.)
 Manhatan= d= |x2-x1|+|y2-y1|
 Euclidean= d= squareroot((x2-x1)2+(y2-y1)2)
 Contd..
 Admissibility of the heuristic function is given as:
 h(n) <= h*(n)
 Here h(n) is heuristic cost, and h*(n) is the
estimated cost.
 Hence heuristic cost should be less than or equal to
the estimated cost.
 Pure Heuristic Search
 The simplest form of heuristic search algorithms
 Expand nodes based on their heuristic value h(n).
 Maintains two lists:
 CLOSED list: contains all of the traversed nodes
 OPEN list: keeps track of the current ‘immediate’ nodes
available for traversal (nodes which have yet not been
expanded)

 The algorithm continues until a goal state is found

 Informed Search Algorithms
 In the informed search we will discuss two main
algorithms:

 Best First Search Algorithm(Greedy search)

 Greedy w.r.t heuristic value
 A* Search Algorithm
 Greedy best-first search algorithm
 In greedy search, we expand the node closest to the goal
node.
 The “closeness” is estimated by a heuristic h(n).
f(n)= h(n).

 Heuristic: A heuristic h is defined as:

 h(n) = Estimated cost from node n to the goal.
 Lower the value of h(n), closer is the node from the goal.
 Strategy: Expand the node closest to the goal state
 i.e. expand the node with a lower h value.
 Implemented by the priority queue.
 Best first search algorithm
 Let OPEN be a priority queue containing initial state.
 LOOP
 If OPEN is empty, stop and return failure
 Else Remove the first node from OPEN with the lowest value of
h(n) and place it in the CLOSED list.
 If node is a goal
 Return the path from initial node to current node
 Else generate all the successors of node n and put newly
generated nodes into OPEN according to their h(n) value

 END LOOP
 Example
For heuristic value, Straight
line distance is calculated
7 D using Euclidean distance
A 25
14 A -> G = 40
11 10 B -> G = 32
C F
B
20 C -> G = 25
8 D -> G = 35
15 G E -> G = 19
10 F -> G = 17
E 9
H H -> G = 10
G -> G = 0
Here numbers mentioned on edges are
Best-First search focuses
just cost to travel from one state to
other rather than heuristic value
only on heuristic values
 D
H=35

G
C
 Path: A>C>F>G
35
7 D A -> G = 40
A 25 B -> G = 32
14 17 C -> G = 25
25 D -> G = 35
11 10
32 C F
20
E -> G = 19
B F -> G = 17
8 0
H -> G = 10
15 G
10 G -> G = 0
E 9
19 H
A C F G E B D

A C F G
 Contd..
 Example solution on board…

 It is important to note that heuristics can sometimes

misguide us.
 E.g. path A -> C -> E-> H -> G has less cost

 Similarly, consider that the maze has fences fixed

along some of the paths through which the smell can
pass.
 Pros and Cons
 Advantages
 Works well with informed search problems, with
fewer steps to reach a goal
 more efficient than BFS and DFS algorithms
 Disadvantages
 Can turn into unguided DFS in the worst case.
 Can get stuck in a loop like DFS.
 This algorithm is not optimal.
 Performance Measurement
 Complete ?
 NO, GBFS can get stuck in loops
 Optimal ?
 NO, heuristic value may misguide.
 Time ?
 The worst case time complexity of Greedy best first search is
O(bm). Simply O(n)
 Space ?
 The worst case space complexity of Greedy best first search
is O(bm).
 Task
 Find the path from S to G using greedy search. The heuristic
values h of each node is given below the name of the node.
 A* Search
 A* (A star) is the most widely known form of Best-First search
 A* Search combines the strengths of uniform-cost search and
greedy search.
 It evaluates nodes by combining g(n) and h(n):
 f(n) = g(n) + h(n)
 Where
 g(n) = cost so far to reach n (backward cost, i.e. UCS cost)
 h(n) = estimated cost to goal from n (i.e. greedy search cost,
forward cost)
 f(n) = estimated total cost of path through n

 Same algorithm as discussed previously just with a difference of f(n)

instead of h(n)
 Contd..
 A* search is optimal only when for all nodes, the
forward cost for a node h(n) underestimates the actual
cost h*(n) to reach the goal. This property
of A* heuristic is called admissibility.
Admissibility

 Strategy: Choose the node with the lowest f(n) value.

 Example
 Find the most cost-effective path to reach from start
state A to final state G using A* Algorithm.
 Solution
 Let’s start with node A. Since A is a starting node, therefore,
the value of g(n) for A is zero and from the graph, we get the
heuristic value of A is 11, therefore
f(n) = g(n) + h(n)
f(n) = 0+ 11 =11
Thus for A, we can write A=11

 Now from A, we can go to point B or point E, so we compute

f(n) for each of them
A→B=2+6=8
A → E = 3 + 7 = 10
 Since the cost for A → B is less, we move forward with this
path and compute the f(n) for the children nodes of B
 Contd..
 Since there is no path between C and G, the heuristic cost is
set infinity or a very high value
A → B → C = (2 + 1) + 99= 102
A → B → G = (2 + 9 ) + 0 = 11

 Here the path A → B → G has the least cost but it is still more
than the cost of A → E, thus we explore this path further
A → E → D = (3 + 6) + 1 = 10
 Comparing the cost of A → E → D with all the paths we got so
far and as this cost is least of all we move forward with this
path. And compute the f(n) for the children of D
A → E → D → G = (3 + 6 + 1) +0 =10
 Contd..
 Now comparing all the paths that lead us to the goal,
we conclude that A → E → D → G is the most cost-
effective path to get from A to G.
 Task

12 11
14 4 B 5
F
S 16
12

3 11 4 G
10 E 5
C
7 2

D
6
 Pros and Cons
 Advantages
 Best algorithm than other search algorithms.
 Optimal and complete.

 Can solve very complex problems.

 Disadvantages
 It does not always produce the shortest path as it mostly
based on heuristics and approximation.
 A* search algorithm has some complexity issues.
 The main drawback of A* is memory requirement as it
keeps all generated nodes in the memory.
 Performance Measurement
 Complete ?
 YES, with finite branching factor and fixed cost
 Optimal ?
 YES, with admissible heuristic search.
 Time ?
 O(bd). depends on heuristic function, and the number of nodes
expanded is exponential to the depth of solution d
 Space ?
 The worst case space complexity of Greedy best first search
is O(bd).
 Task
 Find optimal solution using greedy and A* search.
END