CS 6511: Artificial Intelligence

Informed Search

Amrinder Arora
[Original version of these slides was created by Dan Klein and Pieter Abbeel for Intro to AI at UC Berkeley. http://ai.berkeley.edu]
 Course Logistics
§ Communication
§ Announcements on Slack
§ Projects
§ Team based contests P2 and P4 have a lot of weightage
§ The class winner gets (objectively) more points than those that are not the winners.

AI-4511/6511 GWU 2
 Course Outline
What is AI

Introduction
Env. and Agent
Types

Other AI Topics (Not Uninformed

included in this class): Search

• Robotics
Informed
• NLP Search
• Machine Learning Making
AI / 6511 / Decisions
• Big Data 4511 Constraint
• AR/VR Satisfaction

• Speech Synthesis
Adversarial
Search

Bayes’ Nets

Reasoning Decision
under Theory
Uncertainty
Markov
AI-4511/6511 GWU Decision 3
Processes
 Today
§ Informed Search
§ Concept of “Direction”
§ In physical space
§ In logical solution space
§ Heuristics
§ Greedy Search
§ A* Search
§ Graph Search
AI-4511/6511 GWU 4
 Recap: Search

AI-4511/6511 GWU 5
 Recap: Search

§ Search problem:
§ States (configurations of the world)
§ Actions and costs
§ Successor function (world dynamics)
§ Start state and goal test

§ Search tree:
§ Nodes: represent plans for reaching states
§ Plans have costs (sum of action costs)

§ Search algorithm:
§ Systematically builds a search tree
§ Chooses an ordering of the fringe (unexplored nodes)
§ Optimal: finds least-cost plans
AI-4511/6511 GWU 6
 Example: Pancake Problem

Cost: Number of pancakes flipped

AI-4511/6511 GWU 7
 Example: Pancake Problem

AI-4511/6511 GWU 8
 Example: Pancake Problem
State space graph with costs as weights

4
2 3
2
3

4
3
4 2

3 2
2
4

3
AI-4511/6511 GWU 9
 General Tree Search You know exactly
where you came
from and how you
got there, but you
have no idea where
you’re going. But,
you’ll know it when
you see it.

Action: flip top two Action: fliptoallreach

Path four goal:
Cost: 2 Cost:
Flip 4 flip three
four,
Total cost: 7

AI-4511/6511 GWU 10
 The One Queue – A Useful Abstraction
§ All these search algorithms are the
same except for fringe strategies
§ Conceptually, all fringes are priority
queues (i.e. collections of nodes with
attached priorities)
§ Practically, for DFS and BFS, you can
avoid the log(n) overhead from an
actual priority queue, by using stacks
and queues
§ Can even code one implementation
that takes a variable queuing object

AI-4511/6511 GWU 11
 Uninformed Search
Uninformed search is
structured search, but
does not have any way of
knowing which way to go. It
is an improvement in terms
of endless looping, but not
intelligent.

AI-4511/6511 GWU 12
 Uniform Cost Search

§ Strategy: expand lowest path cost (BFS, but … c£1

considers weights/costs of edges) c£2
c£3

§ The good: UCS is complete and optimal!

§ The bad:
§ Explores options in every “direction”
Start Goal
§ No information about goal location

AI-4511/6511 GWU 13
 Informed Search
§ Concept of “direction”

§ Direction is one kind of information, not the only kind of

information.

AI-4511/6511 GWU 14
 How do we use Direction/Information
§ We “hear” the direction where the sound is coming from, based
on the distance (the time) it takes for the sound to reach the two
ears.
§ “Studies of barn owls offer insight into just how the brain combines
acoustic signals from two sides of the head into a single spatial
perception”
§ https://www.scientificamerican.com/article/listening-with-two-ears-
2006-09/

AI-4511/6511 GWU 15
 Search Heuristics
§ A heuristic is:
§ A function that estimates how close a state is to a goal
§ Designed for a particular search problem
§ Examples: Manhattan distance, Euclidean distance for
pathing

10

5
11.2

AI-4511/6511 GWU 16
 Heuristic Function: Example 1
Heuristic: the number of the largest pancake that is still out of place
3
4
h(x)

3
4
3 0
4
4 3
4
4 2
3
AI-4511/6511 GWU 17
 Heuristic Function: Example 2
§ Euclidean (Flying) distance

AI-4511/6511 GWU 18
 Greedy (Informed) Search
b
§ Strategy: expand a node that you think is …
closest to a goal state
§ Heuristic: estimate of distance to nearest goal for
each state

§ A common case:
b
§ Best-first takes you straight to the (wrong) goal …

§ Worst-case: like a badly-guided DFS

AI-4511/6511 GWU 19
 Comparing UCS and Greedy
§ Uniform-cost orders by path cost, or backward cost g(n)
§ Greedy orders by goal proximity, or forward cost h(n)
8 g=0
S h=6
h=1 g=1
e a
1 h=5

1 3 2 g=2 g=9
S a d G
h=6 b d g=4 e h=1
h=6 h=5 h=2
1 h=2 h=0
1 g=3 g=6
c b g = 10
h=7 c G h=0 d
h=2
h=7 h=6
g = 12
G h=0
AI-4511/6511 GWU 20
Example: Teg Grenager
 Pros and Cons (and What to Do)
§ Neither approach is bad
§ UCS doesn’t take direction into account
§ Greedy doesn’t take the past covered distance into account.

§ We should combine the two ideas though.

AI-4511/6511 GWU 21
 A*: Combining UCS and Greedy
§ Uniform-cost orders by path cost, or backward cost g(n)
§ Greedy orders by goal proximity, or forward cost h(n)

§ A* Search orders by the sum: f(n) = g(n) + h(n)

§ Why is this a good strategy?

AI-4511/6511 GWU 22
Example: Teg Grenager
 When should A* terminate?

§ Should we stop when we enqueue a goal?

h=2

2 A 2

S h=3 h=0 G

2 B 3
h=1

§ No: only stop when we dequeue a goal

AI-4511/6511 GWU 23
 Admissible Heuristics

AI-4511/6511 GWU 24
 Idea: Admissibility

Inadmissible (pessimistic) heuristics break Admissible (optimistic) heuristics slow down

optimality by trapping good plans on the bad plans but never outweigh true costs
fringe
AI-4511/6511 GWU 25
 Admissible Heuristics
§ A heuristic h is admissible (optimistic) if:

where is the true cost to a nearest goal

§ Examples:
4
15
§ Coming up with admissible heuristics is most of what’s involved
in using A* in practice.
AI-4511/6511 GWU 26
 Optimality of A* Tree Search

AI-4511/6511 GWU 27
 Optimality of A* Tree Search
Assume:
§ A is an optimal goal node …
§ B is a suboptimal goal node
§ h(x) is an admissible heuristic

Claim:
§ A will exit the fringe before B

AI-4511/6511 GWU 28
 Optimality of A* Tree Search: Blocking
Proof:
…
§ Imagine B is on the fringe
§ Some ancestor n of A is on the
fringe, too (maybe A!)
§ Claim: n will be expanded before B
1. f(n) is less or equal to f(A)

Definition of f-cost
Admissibility of h
h = 0 at a goal
AI-4511/6511 GWU 29
 Optimality of A* Tree Search: Blocking
Proof:
…
§ Imagine B is on the fringe
§ Some ancestor n of A is on the
fringe, too (maybe A!)
§ Claim: n will be expanded before B
1. f(n) is less or equal to f(A)
2. f(A) is less than f(B)

B is suboptimal
h = 0 at a goal

AI-4511/6511 GWU 30
 Optimality of A* Tree Search: Blocking
Proof:
…
§ Imagine B is on the fringe
§ Some ancestor n of A is on the
fringe, too (maybe A!)
§ Claim: n will be expanded before B
1. f(n) is less or equal to f(A)
2. f(A) is less than f(B)
3. n expands before B
§ All ancestors of A expand before B
§ A expands before B
§ A* search is optimal
AI-4511/6511 GWU 31
 Properties of A*

AI-4511/6511 GWU 32
 Properties of A*

Uniform-Cost A*

b b
… …

AI-4511/6511 GWU 33
 UCS vs. A* Contours

§ Uniform-cost expands equally in all

“directions”
Start Goal

§ A* expands mainly toward the goal,

but does hedge its bets to ensure
optimality Start Goal

AI-4511/6511 GWU 34
 Comparison

Greedy Uniform Cost A*

AI-4511/6511 GWU 35
 A* Applications
§ Video games
§ Pathing / routing problems
§ Resource planning problems
§ Robot motion planning
§ Language analysis
§ Machine translation
§ Speech recognition
§ …
AI-4511/6511 GWU 36
 Creating Heuristics
§ The main point of improvement!!

AI-4511/6511 GWU 37
 Creating Admissible Heuristics
§ Most of the work in solving hard search problems optimally is in coming up
with admissible heuristics

§ Often, admissible heuristics are solutions to relaxed problems, where new

actions are available

366
15

§ Inadmissible heuristics are often useful too

AI-4511/6511 GWU 38
 Example: 8 Puzzle

Start State Actions Goal State

§ What are the states?

§ How many states?
§ What are the actions?
§ How many successors from the start state?
§ What should the costs be?
AI-4511/6511 GWU 39
 8 Puzzle I
§ Heuristic: Number of tiles misplaced
§ Why is it admissible?
§ h(start) = 8
§ This is a relaxed-problem heuristic
Start State Goal State

Average nodes expanded when

the optimal path has…
…4 steps …8 steps …12 steps
UCS 112 6,300 3.6 x 106
A* TILES 13 39 227
AI-4511/6511 GWU 40
 8 Puzzle II

§ What if we had an easier 8-puzzle where

any tile could slide any direction at any
time, ignoring other tiles?

§ Total Manhattan distance

Start State Goal State

§ Why is it admissible? Average nodes expanded when the

optimal path has…
§ h(start) = 3 + 1 + 2 + … = 18
…4 steps …8 steps …12 steps
A* TILES 13 39 227
A* MANHATTAN 12 25 73
AI-4511/6511 GWU 41
 8 Puzzle III
§ How about using the actual cost as a heuristic?
§ Would it be admissible?
§ Would we save on nodes expanded?
§ What’s wrong with it?

§ With A*: a trade-off between quality of estimate and work per node
§ As heuristics get closer to the true cost, you will expand fewer nodes but usually
do more work per node to compute the heuristic itself

AI-4511/6511 GWU 42
 Semi-Lattice of Heuristics

AI-4511/6511 GWU 43
 Trivial Heuristics, Dominance

§ Dominance: ha ≥ hc if

§ Heuristics form a semi-lattice:

§ Max of admissible heuristics is admissible

§ Trivial heuristics
§ Bottom of lattice is the zero heuristic (what
does this give us?)
§ Top of lattice is the exact heuristic
AI-4511/6511 GWU 44
 8 Puzzle, Beyond Manhattan Distance
§ Even if using A* (or other algorithms such as IDA*) along with a
heuristic such as Manhattan Distance, larger puzzles, such as 24
puzzles are still untenable.
§ This is because Manhattan Distance does not take into account
linear conflicts. For example:
• Manhattan Distance is 4, but tiles 1 and 3
interfere with each other.
• [Hansson, Mayer, and Yung, 1991] show that
given two tiles in their goal row, but reversed in
position, additional vertical moves can be added
to Manhattan distance.
• So, 4 + 2 = 6 in this case.
AI-4511/6511 GWU 45
 8 Puzzle, Beyond Manhattan Distance
§ So, using A* and using the New Heuristic (Manhattan + Vertical
Moves), larger puzzles, such as 24 puzzles are still untenable.
§ We need to use a pattern database.
§ A pattern database is a complete set of such positions, with
associated number of moves
§ A 7-tile pattern database for the Fifteen Puzzle contains 519
million entries.

AI-4511/6511 GWU 46
 Using Pattern Database and Semi-Lattice
§ From a given 15-puzzle we may recognize two different patterns
using two different sets of tiles.
§ If one pattern suggests a distance of 20 and the other pattern
suggests a distance of 30, we can take the maximum (30).

AI-4511/6511 GWU 47
 Graph Search

AI-4511/6511 GWU 48
 Tree Search: Extra Work!
§ Failure to detect repeated states can cause exponentially more work.

State Graph Search Tree

AI-4511/6511 GWU 49
 Graph Search
§ In BFS, for example, we shouldn’t bother expanding the circled nodes (why?)

d e p

b c e h r q

a a h r p q f

p q f q c G

q c G a

AI-4511/6511 GWU 50
 Graph Search
§ Idea: never expand a state twice

§ How to implement:
§ Tree search + set of expanded states (“closed set”)
§ Expand the search tree node-by-node, but…
§ Before expanding a node, check to make sure its state has never been
expanded before
§ If not new, skip it, if new add to closed set

§ Important: store the closed set as a set, not a list

§ Can graph search wreck completeness? Why/why not?

§ How about optimality?

AI-4511/6511 GWU 51
 A* Graph Search Gone Wrong?
State space graph Search tree

A S (0+2)
1
1
S h=4
C
h=1 A (1+4) B (1+1)
h=2 1
2
3 C (2+1) C (3+1)
B

h=1 G (5+0) G (6+0)

G

h=0
AI-4511/6511 GWU 52
 Consistency of Heuristics
§ Main idea: estimated heuristic costs ≤ actual costs
§ Admissibility: heuristic cost ≤ actual cost to goal
A
1 h(A) ≤ actual cost from A to G
h=4 C h=1 § Consistency: heuristic “arc” cost ≤ actual cost for each arc
h=2
h(A) – h(C) ≤ cost(A to C)
3
§ Consequences of consistency:
§ The f value along a path never decreases
G h(A) ≤ cost(A to C) + h(C)

§ A* graph search is optimal

AI-4511/6511 GWU 53
 Optimality of A* Graph Search

AI-4511/6511 GWU 54
 Optimality of A* Graph Search

§ Sketch: consider what A* does with a

consistent heuristic:
… f£1
§ Fact 1: In tree search, A* expands nodes in
f£2
increasing total f value (f-contours)
f£3
§ Fact 2: For every state s, nodes that reach
s optimally are expanded before nodes
that reach s suboptimally

§ Result: A* graph search is optimal

AI-4511/6511 GWU 55
 Optimality
§ Tree search:
§ A* is optimal if heuristic is admissible
§ UCS is a special case (h = 0)

§ Graph search:
§ A* optimal if heuristic is consistent
§ UCS optimal (h = 0 is consistent)

§ Consistency implies admissibility

§ In general, most natural admissible heuristics

tend to be consistent, especially if from
relaxed problems

AI-4511/6511 GWU 56
 A*: Summary
§ A* uses both backward costs and (estimates of) forward costs

§ A* is optimal with admissible / consistent heuristics

§ Heuristic design is key: often use relaxed problems

AI-4511/6511 GWU 57
 Tree Search Pseudo-Code

AI-4511/6511 GWU 58
 Graph Search Pseudo-Code

AI-4511/6511 GWU 59
 Optimality of A* Graph Search

§ Consider what A* does:

§ Expands nodes in increasing total f value (f-contours)
Reminder: f(n) = g(n) + h(n) = cost to n + heuristic
§ Proof idea: the optimal goal(s) have the lowest f value, so
it must get expanded first

… f£1
There’s a problem with this f£2
argument. What are we assuming
f£3
is true?

AI-4511/6511 GWU 60
 Optimality of A* Graph Search

Proof:
§ New possible problem: some n on path to G*
isn’t in queue when we need it, because some
worse n’ for the same state dequeued and
expanded first (disaster!)
§ Take the highest such n in tree
§ Let p be the ancestor of n that was on the
queue when n’ was popped
§ f(p) < f(n) because of consistency
§ f(n) < f(n’) because n’ is suboptimal
§ p would have been expanded before n’
§ Contradiction!

AI-4511/6511 GWU 61