Part 3.

3: Informed (Heuristic)
Search
Concept
Informed search methods
Search with Domain Knowledge added


Uninformed (blind) searches are normally very
inefficient

Adding domain knowledge can improve the search
process.

Kaburu.M.M 2
cisy422_bbit333
Concept of informed (heuristic) search


Heuristic (informed) search -> explore the node that
is most “likely” to be the nearest to a goal state.

There is no guarantee that the heuristic provided
most “likely” node will get you closer to a goal state
than any other.

Kaburu.M.M 3
cisy422_bbit333
Knowledge/info


Example:
 City streets

You are at GPO and would like to go the Old Nation
Roundabout to get a matatu - What path will you take?
Is it the shortest?
 Parallel streets on all sides

Kaburu.M.M 4
cisy422_bbit333
Knowledge/info


Example:
 An office in a city building

Find me in Office door number N311 e.g. KICC, NSSF,
KEMU Main Campus, Student’s Center?
 Wing, floor, left/right

 Some streets in Nairobi are named in alphabetical

order

Kaburu.M.M 5
cisy422_bbit333
Knowledge/info


Example:
 Visit the doctor

Symptoms: fever, nausea, headache, …
 Leading questions: how long?, traveled?, … (Malaria,
typhoid, meningitis, flu,..)
 x Blood test, y test,...

Kaburu.M.M 6
cisy422_bbit333
Knowledge/info


Example:
 Climbing a hill in thick Fog

Heuristic function: check the change in altitude in 4
directions; the strongest increase is the direction in
which to move next.

Kaburu.M.M 7
cisy422_bbit333
Heuristic Searches - Characteristics


Has some domain knowledge

Usually more efficient than blind searches

Also called informed search

Heuristic searches work by deciding which is the
next best node to expand (there is no guarantee that
it is the best node)

Kaburu.M.M 8
cisy422_bbit333
Heuristic Searches - Why Use?

It may be too resource intensive (both time and
space) to use a blind search

Even if a blind search will work we may want a more
efficient search method

Kaburu.M.M 9
cisy422_bbit333
Part 3.3: Informed (Heuristic)
Search
Concept
Informed search methods

Kaburu.M.M 10
Heuristic Search Methods


Methods that use a heuristic function to provide
specific knowledge about the problem:

Heuristic Functions

Hill climbing

Greedy search

A* search algorithm

Kaburu.M.M 11
cisy422_bbit333
Heuristic Functions

To further improve the quality of the previous
methods, we need to include problem-specific
knowledge on the problem.
 How can this be done in such a way that the
algorithms remain generally applicable ???

 HEURISTIC FUNCTIONS:
 f: States --> Numbers
 f(T) : expresses the quality of the state T
 allow to express problem-specific knowledge in
the search method algorithm
Kaburu.M.M 12
cisy422_bbit333
 Example 1: road map

Imagine the problem of finding a route on a road
map and that the NET below is the road map:
4 4
3 A B C
S 5 5 G
4 D E F 3
2 4
 Define f(T) = the straight-line distance from T to G
A 10.4 B 6.7 C 4
11 The estimate
S
G can be wrong!
8.9
D E 6.9 F 3
Kaburu.M.M 13
cisy422_bbit333
 Example 2: 8-puzzle

f1(T) = the number correctly placed tiles on the
board: 1 3 2
f1 8 4 =4
5 6 7

 f2(T) = number or incorrectly placed tiles on board:

 gives (rough!) estimate of how far we are from goal

1 3 2
f2 8 4 =4
5 6 7
Most often, ‘distance to goal’ heuristics are more useful !
Kaburu.M.M 14
cisy422_bbit333
Example 2: 8-puzzle Manhattan distance


f3(T) = the sum of ( the horizontal + vertical distance
that each tile is away from its final destination):
 gives a better estimate of distance from the goal
node

1 3 2
f2 8 4 =1+1+2+2=6
5 6 7

Kaburu.M.M 15
cisy422_bbit333
Hill climbing


A basic heuristic search method:
 depth-first + heuristic

Kaburu.M.M 16
cisy422_bbit333
 Hill climbing_1
 Example: using the straight-line distance:

S

Perform depth-first,
10.4 8.9 BUT:
A D 
instead of left-to-right
selection,
10.4 6.9
A E

FIRST select the child
with the best heuristic
value
6.7 B F
3.0

Kaburu.M.M 17
cisy422_bbit333
Hill climbing_2
 Inspiring Example: climbing a hill in the fog.
 Heuristic function: check the change in altitude

in 4 directions: the strongest increase is the

direction in which to move next.

 Is identical to Hill climbing_1, except for dropping

the backtracking.
 Produces a number of classical problems:

Kaburu.M.M 18
cisy422_bbit333
 Problems with
Hill climbing_2:
Foothills: Plateaus

Local
maximum

Ridges

Kaburu.M.M 19
cisy422_bbit333
 Comments:

Foothills are local minima: hill climbing_2 can’t
detect the difference.

Plateaus don’t allow you to progress in any
direction.

 Foothills and plateaus require random jumps to

be combined with the hill climbing algorithm.


Ridges neither: the directions you have fixed in
advance all move downwards for this surface.
 Ridges require new rules, more directly targeted
to the goal, to beKaburu.M.M
introduced (new directions to 20

move) . cisy422_bbit333
 Local search
S

Hill climbing_2 is an example of local
search. A

In local search, we only keep track of 1 B
path and use it to compute a new path in
the next step. C
D
 QUEUE is always of the form:

( p)
 Another example:
 MINIMAL COST search: A
3 4
 If p is the current path:
B C
 the next path extends p by adding 2 3 5
the node with the smallest cost from D E F
the endpoint of p Kaburu.M.M 21
cisy422_bbit333
 Heuristic Searches - Greedy search


So named as it takes the biggest “bite” it can out of
the problem.
That is, it seeks to minimise the estimated cost to the
goal by expanding the node estimated to be closest
to the goal state

in other words,

Always expand the heuristically best nodes first.

Kaburu.M.M 22
cisy422_bbit333
 Greedy search, or
Heuristic best-first search:

S 30

10 40
1 20 
At each step,

3 select the node

with the best (in
35 15 this case:
lowest) heuristic
2 27 18 value.

25 45 The ‘open’
nodes
Kaburu.M.M 23
cisy422_bbit333
 Greedy search algorithm:
1. QUEUE <-- path only containing the root;

2. WHILE QUEUE is not empty

AND goal is not reached

DO remove the first path from the QUEUE;

create new paths (to all children);
reject the new paths with loops;
add the new paths and sort the entire QUEUE;
(HEURISTIC)
3. IF goal reached
THEN success;
ELSE failure; Kaburu.M.M 24
cisy422_bbit333
Heuristic Searches - Greedy Search

It is only concerned with short term aims

It is possible to get stuck in an infinite loop, unless
you check for repeated states

It is not optimal

It is not complete

Time and space complexity is O(Bm); where m is the depth of the

search tree

Kaburu.M.M 25
cisy422_bbit333
 Heuristic Searches - A* Algorithm

A combination of Greedy search and Uniform cost
search. - to compliment one another

This search method minimises the cost to the goal
using an heuristic function, h(n). Greedy search can
considerably cut the search time but it is neither
optimal nor complete. By comparison uniform cost
search minimises the cost of the path so far, g(n).
Uniform cost search is both optimal and complete
but can be very inefficient.

Kaburu.M.M 26
cisy422_bbit333
 Road map example:
Re-introduce the costs of paths in the NET

A 4 B 4 C
3
S 5 5
G
4 D 2 E F 3
4
3 S 4
4 A D
5 5 2
B D A E
4 5 2 4 5 4
C E E B B F
2 4 5 4 4 5 4 4 3
D F B F C E A C G
3 4 3 4
G C G F
Kaburu.M.M 3 27
cisy422_bbit333 G
A look at Uniform cost search= uniformed best-first

S
3 4

A 3 5 D4 At each step,
4


5 2 select the
B 7 D 8 A 9 E 6 node with the
4 5 5 4 lowest
5 4
accumulated
C 11 E 12 E 13 B 13 B 11 F 10 cost.
3
D F B F C E A C G 13

G C G F
G

Kaburu.M.M 28
cisy422_bbit333
Now incorporate heuristic estimates

Replace the ‘accumulated cost’ in the ‘Uniform
cost search’ by a function:
f(path) = cost(path) + h(endpoint_path)

fn g(n) h(n)
 where:
cost(path) = the accumulated cost of the partial path
h(T) = a heuristic estimate of the cost remaining
from T to a goal node

f(path) = an estimate of the cost of a path extending

the current path to reach a goal.
fn =Kaburu.M.M
g(n) + h(n) 29
cisy422_bbit333
 Heuristic Searches - A* Algorithm

Combines the cost so far and the heuristic estimated
cost to the goal. That is fn = g(n) + h(n)
This gives us estimated cost of the cheapest solution
through path n

It can be proved to be optimal and complete
providing that the heuristic is admissible.
That is the heuristic must never over estimate the
cost to reach the goal.

But, the number of nodes that have to be searched
still grows exponentially

Kaburu.M.M 30
cisy422_bbit333
 Example: road map

Imagine the problem of finding a route on a road
map. The paths distances between nodes define
g(n): 4 4
3 A B C
S 5 5 G
4 D E F 3
2 4
 Define h(n) = the straight-line distance from node to
G 10.4
A B 6.7 C 4
11
S
G
8.9
D E 6.9 F 3
Kaburu.M.M 31
cisy422_bbit333
 S
3 + 10.4 = 13.4 A D 4 + 8.9 = 12.9

S
A 13.4 D
9 + 10.4 = 19.4 A E 6 + 6.9 = 12.9
X
S
A 13.4 D
A E
X B F 10 + 3.0 = 13.0
11 + 6.7 = 17.7
S
A 13.4 D
A
X
E STOP!
17.7 B F
Kaburu.M.M 32
cisy422_bbit333
G 13 + 0.0 = 13.0
 Heuristic Searches - A* Algorithm for 8-Puzzle
Example
5 4 1 2 3
6 1 8 8 4
7 3 2 7 6 5

Initial State Goal State

 Typical solution is about twenty steps

 Branching factor is approximately three. Therefore a complete

search would need to search 320 states. But by keeping track of
repeated states we would only need to search 9! (362,880) states

 But even this is a lot (imagine having all these in memory)

 Our aim is to develop a heuristic that does not over estimate (it is
admissible) so that we can use A* to find the optimal solution
Kaburu.M.M 33
cisy422_bbit333
Heuristic Searches - Possible Heuristics

H1 = the number of tiles that are in the wrong

position (=7)

H2 = the sum of the distances of the tiles from their

goal positions using the Manhattan Distance (=18)

Both are admissible but which one is best?

Kaburu.M.M 34
cisy422_bbit333
Test from 100 runs with varying solution depths
Search
Cost
Depth IDS A*(h1) A*(h2)
2 10 6 6
4 112 13 12
6 680 20 18
8 6384 39 25
10 47127 93 39
12 364404 227 73
14 3473941 539 113
16 1301 211
18 3056 363
20 7276 676
22 18094 1219
24 39135 1641
H2 looks better as fewer nodes are expanded. But why?
Kaburu.M.M 35
cisy422_bbit333
 Effective Branching Factor
Search EBF
Cost
Depth IDS A*(h1) A*(h2) IDS A*(h1) A*(h2)
2 10 6 6 2.45 1.79 1.79
4 112 13 12 2.87 1.48 1.45
6 680 20 18 2.73 1.34 1.30
8 6384 39 25 2.80 1.33 1.24
10 47127 93 39 2.79 1.38 1.22
12 364404 227 73 2.78 1.42 1.24
14 3473941 539 113 2.83 1.44 1.23

 H2 has a lower branching factor and so fewer nodes are expanded

 Therefore, one way to measure the quality of a heuristic is to find
its average branching factor
 H2 has a lower EBF and is therefore the better heuristic

Kaburu.M.M 36
cisy422_bbit333
Summary


Blind Search is very expensive

A Search with info (heuristics) is far better

Info can be difficult to incorporate

The more info, the better

Kaburu.M.M 37
cisy422_bbit333