Informed search algorithms

Outline
 Best-first search
 Greedy best-first search
 A* search
 Heuristics
 Local search algorithms
 Hill-climbing search
 Simulated annealing search
 Local beam search
 Genetic algorithms
Best-first search
 Idea: use an evaluation function f(n) for each node
 f(n) provides an estimate for the total cost.
 Expand the node n with smallest f(n).

 Implementation:
Order the nodes in fringe increasing order of cost.

 Special cases:
 greedy best-first search
 A* search
Romania with straight-line dist.
 Greedy best-first search
 f(n) = estimate of cost from n to goal
 e.g., f(n) = straight-line distance from n
to Bucharest
 Greedy best-first search expands the
node that appears to be closest to goal.
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!
A* search
 Idea: avoid expanding paths that are already
expensive
 Evaluation function f(n) = g(n) + h(n)
 g(n) = cost so far to reach n
 h(n) = estimated cost from n to goal
 f(n) = estimated total cost of path through n to
goal
 Best First search has f(n)=h(n)
 Uniform Cost search has f(n)=g(n)
A* search example
A search example
*
A* search example
A* search example
A* search example
A* search example
 Admissible heuristics
 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.
 An admissible heuristic never overestimates the cost to
reach the goal, i.e., it is optimistic
 Example: hSLD(n) (never overestimates the actual road
distance)
 Theorem: If h(n) is admissible, A* using TREE-SEARCH
is optimal
 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.

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

Consistent heuristics
 A heuristic is consistent if for every node n, every successor n' of n
generated by any action a,

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: keeps all checked nodes

If h(n) is consistent, A* using GRAPH-SEARCH is optimal
in memory to avoid repeated
states
Optimality of A*
 A* expands nodes in order of increasing f value
 Gradually adds "f-contours" of nodes
 Contour i contains all nodes with f≤fi where fi < fi+1
 Properties of A*
 Complete? Yes (unless there are infinitely many
nodes with f ≤ f(G) , i.e. path-cost > ε)
 Time/Space? Exponential b d
* *
except if: | h (n )  h (n ) | O (log h (n ))

 Optimal? Yes
 Optimally Efficient: Yes (no algorithm with the
same heuristic is guaranteed to expand fewer nodes)
 straight-line distances
6 1
3 A D F 1 h(S-G)=10
h(A-G)=7
2 4 8
S B E G h(D-G)=1
h(F-G)=1
1 h(B-G)=10
20
C h(E-G)=8
h(C-G)=20

try yourself

The graph above shows the step-costs for different paths going from the start (S) to
the goal (G). On the right you find the straight-line distances.

1. Draw the search tree for this problem. Avoid repeated states.

2. Give the order in which the tree is searched (e.g. S-C-B...-G) for A* search.
Use the straight-line dist. as a heuristic function, i.e. h=SLD,
and indicate for each node visited what the value for the evaluation function, f, is.
 Memory Bounded Heuristic
Search: Recursive BFS
 How can we solve the memory problem for
A* search?
 Idea: Try something like depth first search,
but let’s not forget everything about the
branches we have partially explored.
 We remember the best f-value we have
found so far in the branch we are deleting.
RBFS:
best alternative
over fringe nodes,
which are not children:
do I want to back up?

RBFS changes its mind

very often in practice.

This is because the

f=g+h become more
accurate (less optimistic)
as we approach the goal.
Hence, higher level nodes
have smaller f-values and
will be explored first.

Problem: We should keep

in memory whatever we can.
 Simple Memory Bounded A*
 This is like A*, but when memory is full we delete the
worst node (largest f-value).
 Like RBFS, we remember the best descendent in the
branch we delete.
 If there is a tie (equal f-values) we delete the oldest
nodes first.
 simple-MBA* finds the optimal reachable solution
given the memory constraint. A Solution is not reachable
 Time can still be exponential. if a single path from root to goal
does not fit into memory
Admissible heuristics
E.g., for the 8-puzzle:
 h1(n) = number of misplaced tiles
 h2(n) = total Manhattan distance
(i.e., no. of squares from desired location of each tile)

 h1(S) = ?
 h2(S) = ?
Admissible heuristics
E.g., for the 8-puzzle:
 h1(n) = number of misplaced tiles
 h2(n) = total Manhattan distance
(i.e., no. of squares from desired location of each tile)

 h1(S) = ? 8
 h2(S) = ? 3+1+2+2+2+3+3+2 = 18
Dominance
 If h2(n) ≥ h1(n) for all n (both admissible)
 then h2 dominates h1
 h2 is better for search: it is guaranteed to expand
less nodes.

 Typical search costs (average number of nodes

expanded):

 d=12 IDS = 3,644,035 nodes

A*(h1) = 227 nodes
A*(h2) = 73 nodes
 d=24 IDS = too many nodes
A*(h1) = 39,135 nodes
A*(h2) = 1,641 nodes
 Relaxed problems
 A problem with fewer restrictions on the actions
is called a relaxed problem
 The cost of an optimal solution to a relaxed
problem is an admissible heuristic for the
original problem
 If the rules of the 8-puzzle are relaxed so that a
tile can move anywhere, then h1(n) gives the
shortest solution
 If the rules are relaxed so that a tile can move
to any adjacent square, then h2(n) gives the
shortest solution
 Local search algorithms
 In many optimization problems, the path to the goal
is irrelevant; the goal state itself is the solution

 State space = set of "complete" configurations

 Find configuration satisfying constraints, e.g., n-
queens
 In such cases, we can use local search algorithms
 keep a single "current" state, try to improve it.
 Very memory efficient (only remember current state)
 Example: n-queens
 Put n queens on an n × n board with no
two queens on the same row, column,
or diagonal

Note that a state cannot be an incomplete configuration with m<n queens

 Hill-climbing search
 Problem: depending on initial state, can get stuck in local
maxima
 Hill-climbing search: 8-queens
problem

Each number indicates h if we move

a queen in its corresponding column

 h = number of pairs of queens that are attacking each other, either

directly or indirectly (h = 17 for the above state)
 Hill-climbing search: 8-queens
problem

 A local minimum with h = 1

Simulated annealing search
 Idea: escape local maxima by allowing some
"bad" moves but gradually decrease their
frequency.

 This is like smoothing the cost landscape.

Properties of simulated
annealing search
 One can prove: If T decreases slowly enough,
then simulated annealing search will find a
global optimum with probability approaching
1 (however, this may take VERY long)

 Widely used in VLSI layout, airline scheduling,

etc.
Local beam search
 Keep track of k states rather than just one.

 Start with k randomly generated states.

 At each iteration, all the successors of all k

states are generated.

 If any one is a goal state, stop; else select the k

best successors from the complete list and
repeat.
Genetic algorithms
 A successor state is generated by combining two parent
states

 Start with k randomly generated states (population)

 A state is represented as a string over a finite alphabet

(often a string of 0s and 1s)

 Evaluation function (fitness function). Higher values for

better states.

 Produce the next generation of states by selection,

crossover, and mutation
fitness:
#non-attacking queens

probability of being
regenerated
in next generation
 Fitness function: number of non-attacking pairs of queens
(min = 0, max = 8 × 7/2 = 28)
 24/(24+23+20+11) = 31%
 23/(24+23+20+11) = 29% etc
Appendix
 Some details of the MBA* next.
SMA* pseudocode (not in 2nd edition 2 of
book)
function SMA*(problem) returns a solution sequence
inputs: problem, a problem
static: Queue, a queue of nodes ordered by f-cost

Queue  MAKE-QUEUE({MAKE-NODE(INITIAL-STATE[problem])})
loop do
if Queue is empty then return failure
n  deepest least-f-cost node in Queue
if GOAL-TEST(n) then return success
s  NEXT-SUCCESSOR(n)
if s is not a goal and is at maximum depth then
f(s)  
else
f(s)  MAX(f(n),g(s)+h(s))
if all of n’s successors have been generated then
update n’s f-cost and those of its ancestors if necessary
if SUCCESSORS(n) all in memory then remove n from Queue
if memory is full then
delete shallowest, highest-f-cost node in Queue
remove it from its parent’s successor list
insert its parent on Queue if necessary
insert s in Queue
end
 Simple Memory-bounded A* (SMA*)
(Example with 3-node memory) maximal depth is 3, since
memory limit is 3. This
Progress of SMA*. Each node is labeled with its current f-cost. branch is now useless.
Values in parentheses show the value of the best forgotten
descendant. best forgotten node
Search space best estimated solution
so far for that node
f = g+h  = goal A
13[15]
A
0+12=12 A A A
12 12
10 8 13
G
B G 13
10+5=15 8+5=13
B B G
10 10 8 16 15
18 H
15 13
20+5=25
C D
16+2=18
H I 
20+0=20 24+0=24
10 10 A A A
8 8 15[15] 15[24] 20[24]
E F J K
A 8
15
30+5=35 30+0=30 24+0=24 24+5=29 G B B
15 20[]
24[]

B G
I D
15 24 C 25
24 20

Algorithm can tell you when best solution found within memory constraint is optimal or not.