ARTIFICIAL INTELLIGENCE

lecture-3
-Abhijit Boruah
DUIET

1
Informed (Heuristic) Search
 Uses problem specific knowledge beyond the definition of the problem itself.

 General approach is best first search where a node is selected for expansion based
on an evaluation function f(n).,which is a cost estimate.

 Node with lowest f(n) is evaluated first. The choice of f determines the search
strategy.

 Most best first search algorithms include a heuristic function as a component of f :

h(n) = estimated cost of the cheapest path from the state at node n to a goal state.

 Special cases:
 Greedy Best first search
 A* search
Greedy Best First Search
 Greedy best-first search expands the node that appears
to be closest to goal.

 Evaluates node just by its heuristic function, i.e.,

f(n) = h(n).

 e.g., fSLD(n) = straight-line distance from n to

Bucharest
Fsld example-from Arad to Bucharest
Greedy best-first search example
Greedy best-first search example
Greedy best-first search example
Greedy best-first search example
Properties of greedy best-first search
 Complete? No – can get stuck in loops.
 Time? O(bm), but a good heuristic can give dramatic
improvement
 Space? O(bm) - keeps all nodes in memory
 Optimal? No
e.g. AradSibiuRimnicu VireaPitestiBucharest is
shorter!

*m is maximum depth of the search space

A* Search
 Idea: avoid expanding paths that are already expensive.

 It is also worth mentioning that many games and web-based maps use this
algorithm to find the shortest path very efficiently .

 Node evaluation by combining g(n): the cost to reach the node and h(n):
the cost(cheapest) to get to goal from the node.

 f(n) = g(n) + h(n) (estimated cost of the cheapest solution through n) .

 Hence expand the node with the lowest value of g(n) + h(n).
A* search Example
A* search example
A* search example
A* search example
A* search example
A* search example
A* search example
Conditions for optimality: Admissibility
and consistency
 First condition for optimality is admissibility.

 An admissible heuristic never overestimates the cost to reach the goal as

h(n) in f(n) = g(n) + h(n). WHY?

 A heuristic h(n) is admissible if for every node n, h(n) ≤ h*(n), where h*(n)
is the true cost to reach the goal state from n.

 Example: hSLD(n) (never overestimates the actual road distance)

 Theorem: If h(n) is admissible, TREE-SEARCH version of A* is

optimal.
Optimality of A* tree search version (proof)
 Suppose some suboptimal goal G2 has been generated and is in the fringe.
Let n be an unexpanded node in the fringe such that n is on a shortest path
to an optimal goal G.

We want to prove:
f(n) < f(G2)
(then A* will prefer n over G2)

 f(G2) = g(G2) since h(G2) = 0

 f(G) = g(G) since h(G) = 0
 g(G2) > g(G) since G2 is suboptimal
 f(G2) > f(G) from above
 Optimality of A* (proof)

 Suppose some suboptimal goal G2 has been generated and is in the fringe. Let n be
an unexpanded node in the fringe such that n is on a shortest path to an optimal goal
G.

 f(G2) > f(G) copied from last slide

 h(n) ≤ h*(n) since h is admissible (under-estimate)
 g(n) + h(n) ≤ g(n) + h*(n) from above
 f(n) ≤ f(G) since g(n)+h(n)=f(n) & g(n)+h*(n)=f(G)
 f(n) < f(G2) from top line.
Hence: n is preferred over G2
Conditions for optimality: Admissibility and
consistency
 Second and more stronger condition for optimality, also called consistency or
monotonicity.
 A heuristic h(n) is consistent if for every node n, every successor n' of n generated
by any action a, the estimate cost of reaching the goal from n is not greater than
step cost of reaching to n’ and estimated cost of reaching goal from n’

h(n) ≤ c(n,a,n') + h(n')

 If h is consistent, we have
f(n') = g(n') + h(n')
= g(n) + c(n,a,n') + h(n')
≥ g(n) + h(n) = f(n)
f(n’) ≥ f(n) It’s the triangle
 i.e., f(n) is non-decreasing along any path. inequality !

 Theorem:
If h(n) is consistent, GRAPH-SEARCH version of A* is optimal
Heuristics
 We can calculate g but how to calculate h ?
 Either calculate the exact value of h (which is certainly time consuming).
             OR
Approximate the value of h using some heuristics (less time consuming).
 Exact Heuristics
1. Pre-compute the distance between each pair of cells before running
the A* Search Algorithm.
2. If there are no blocked cells/obstacles then we can just find the
exact value of h without any pre-computation using the distance
formula/Euclidean Distance

Video A* Search
Heuristics
 Approximation Heuristics
 Manhattan Distance:  sum of absolute values of differences in the goal’s x and y
coordinates and the current cell’s x and y coordinates respectively, i.e., h = abs
(current_cell.x – goal.x) + abs (current_cell.y – goal.y)
When to use this heuristic? – When we are allowed to move only in four direction
only (right, left, top, bottom)

 Diagonal Distance- maximum of absolute values of differences in the goal’s x and y

coordinates and the current cell’s x and y coordinates respectively, i.e., h = max
{ abs(current_cell.x – goal.x), abs(current_cell.y – goal.y) } . Use it when we are
allowed to move in eight directions only (similar to a move of a King in Chess).

 Euclidean Distance- distance between the current cell and the goal cell using the
distance formula h = sqrt ( (current_cell.x – goal.x) 2 + (current_cell.y – goal.y)2). Use it
when we are allowed to move in any directions.
Properties of A* ( home task)
If C* is the cost of the optimal solution path, then
 A* expands all nodes with f(n) < C*
 A* then might expand some nodes in the goal contour (where f(n) = C*)
before selecting a goal node.
 Completeness? – requires there to be only finitely many nodes with
cost less than or equal to C*.
 Time/Space complexities?
 Time : O(b∆);for constant steps O(bεd ), ∆ is absolute error (h*-h) and ε is relative error
(h*-h)/h*.
 Memory inefficient, runs out of space, not practical for large-scale problems.
 Go for memory bounded heuristic search (eg. Recursive BFS, MA*)
 Optimality? – optimally efficient for any given consistent heuristics
Generation of Heuristic Functions
(Self Study)
 Simple Memory Bound A* search(SMA*) ( go to link
https://userweb.cs.txstate.edu/~ma04/files/CS5346/SMA%20search.pdf )
 Generating admissible heuristics from relaxed problems.
 Generating admissible heuristics from sub problems.
 Learning heuristics from experience.
Assignments
1. Prove each of the following statements:
a) Breadth-first search is a special case of uniform-cost search.
b) Breadth-first search, depth-first search, and uniform-cost search are
special cases of best-first search.
c) Uniform-cost search is a special case of A* search.

2. The heuristic path algorithm is a best-first search in which the objective

function is f(n) = (2 - w)g(n) + wh(n). For what values of w is this
algorithm guaranteed to be optimal? (You may assume that h is
admissible.) What kind of search does this perform when w = O? When
w = l? When w = 2?