COSE361 Artificial Intelligence

Search (2)

[Adapted from the original slides from CS188 Intro to AI at UC Berkeley]

 Review
▪ Agents that Plan Ahead
▪ Reflex Agents
▪ Planning Agents
▪ Search Problems
▪ State Space Graphs
▪ Search Trees
▪ Uninformed Search Methods
▪ Depth-first Search
▪ Breadth-first Search
 Outline
▪ Uninformed Search Methods
▪ Depth-First Search
▪ Breadth-First Search
▪ Iterative Deepening Search
▪ Uniform-Cost Search
▪ Informed (Heuristic) Search
 Iterative Deepening Search
▪ Combining the best of DFS and BFS

Search takes time O(bm) space O(bm) Search takes time O(bs) space O(bs)

May get stuck on infinite paths Guaranteed to find a solution if one exists
Finds the “leftmost” solution Finds the “shallowest” solution
 Iterative Deepening Search
▪ Idea: DFS’s space advantage + BFS’s time/shallow-solution advantages
▪ For every depth limit, perform DFS until the depth limit is reached
▪ Run a DFS with depth limit 1. If no solution…
▪ Run a DFS with depth limit 2. If no solution…
▪ Run a DFS with depth limit 3. …..

b
…
 Example: Iterative Deepening Search
▪ Build search trees for depth = 0, 1, 2 on the search graph below
▪ Starting from S (a root node, depth = 0)
▪ Quiz: what is the total number of nodes visited?

a G
b c

e
d f
S h
p r
q
 Iterative Deepening Search
▪ Idea: DFS’s space advantage + BFS’s time/shallow-solution advantages
▪ For every depth limit, perform DFS until the depth limit is reached

▪ Isn’t that wastefully redundant?

b = 10
▪ No, generally most work happens in the lowest level searched …
▪ # of nodes grows exponentially
s=5
▪ Example: branching factor 10, solution at depth 5:
▪ Number of nodes visited?
▪ BFS: 10 + 100 + 1,000 + 10,000 + 100,000 = 111,110
▪ IDS: 50 + 400 + 3,000 + 20,000 + 100,000 = 123,450
 Iterative Deepening Search
▪ Time Complexity
▪ O(bs), exponential in s, same as BFS

▪ Space Complexity b
…
▪ O(bs), linear in s, similar to DFS
s tiers

▪ Completeness
▪ Yes, same as BFS

▪ Optimality
▪ No, only the shallowest solution, same as BFS
▪ (but Yes if costs are all 1)
 Quiz
▪ Find the shortest path (in terms of # of nodes in the path) from
Start node to Goal node

a G

b c

e
d
f
S h

p r
q
 Cost-Sensitive Search

a GOAL
2 2
b c
3
2
1 8
2 e
Cost: 3 d
f
9 8 2
START h
1 4 2

p 4 r
15
q

▪ BFS finds the shortest path in terms of number of actions, but not the least-cost path
▪ We will now cover a similar algorithm which does find the least-cost path
Uniform Cost Search
 Uniform Cost Search (a.k.a. Dijkstra’s algorithm)
2 a G
Strategy: expand a cheapest node first b c
1 8 2
2 e
Fringe is a priority queue 3 d f
9 2
S h 8
(i.e., ordered by cumulative cost) 1
1 p q r
15

S 0

d 3 e 9 p 1

b 4 c e 5 h 17 r 11 q 16
11
Cost a 6 a h 13 r 7 p q f
contours
p q f 8 q c G

q 11 c G 10 a

a
 Uniform Cost Search (UCS) Properties
▪ What nodes does UCS expand?
▪ Processes all nodes with cost less than cheapest solution!
b c1
▪ If that solution costs C* and arcs cost at least  , then the …
“effective depth” is roughly C*/ C*/ “tiers”
c2

▪ Takes time O(b ) (exponential in effective depth)
C*/
c3

▪ How much space does the fringe take?

▪ Has roughly the last tier, so O(bC*/)

▪ Is it complete?
▪ Yes, assuming best solution has a finite cost and minimum arc
cost is positive

▪ Is it optimal?
▪ Yes
 Uniform Cost Search Issues
▪ UCS explores increasing cost contours c1
…
c2
c3
▪ The good:
▪ UCS is complete and optimal!

▪ The bad:
▪ Explores options in every “direction”
▪ No information about goal location Start Goal
▪ How can we fix that?
Video of Demo Empty UCS
Video of Demo Contours UCS Pacman Small Maze
Video of Demo Maze with Deep/Shallow Water --- DFS, BFS, or UCS?
 Comparing Uninformed Search Algorithms

▪ All these search algorithms are the same except for fringe strategies
▪ Conceptually, all fringes are priority queues (i.e., collections of nodes with priorities)
▪ Practically, for DFS and BFS, you can use LIFO stacks and FIFO queues
 Summary

▪ Search problem:
▪ States (configurations of the world)
▪ Successor function (world dynamics)
▪ Actions and costs
▪ 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)
▪ Finds least-cost plans
Example: Pancake Problem

Cost? Action? State space size?

 Example: Pancake Problem
Partial state space graph with costs (# of pancakes flipped)

4
2 3
2
3

4
3
4 2

3 2
2
4

3
 General Tree Search

Action: flip top two Action: fliptoallreach

Path four goal:
Cost: 2 Cost:
Flip 4 flip three
four,
Total cost: 7
 Outline
▪ Uninformed Search Methods
▪ Depth-First Search
▪ Breadth-First Search
▪ Iterative Deepening Search
▪ Uniform-Cost Search
▪ Informed (Heuristic) Search
▪ Heuristics
▪ Greedy Search
 Informed Search
▪ Uninformed search
▪ Consider every state to be the same
▪ Does not know which state is closer to the goal state
▪ May not find the optimal solution

▪ Informed (with heuristic) search

▪ Estimate how close the state is to the goal
▪ c.f., heuristic (in computing): finding a solution by trying different actions (with soft
rules) to see if they produce the result that is wanted, rather than using strict rules
▪ Try to find the optimal solution with the estimation by intuitive clues
 Search Heuristics
▪ A search heuristic is:
▪ A function h(x) that estimates how close a state x is to a goal G
▪ Designed for a particular search problem
▪ Examples: Manhattan/Euclidean distance for pathing

10

5
11.2
Example: Heuristic Function for Traveling
h(x)
Quiz: Heuristic Function for Pancake
Slido.com
What could be the heuristic function for the pancake problem? → Code: 2650956

Goal state
 Quiz: Heuristic Function for Pancake
Example heuristic: the number (ID) of the largest pancake that is still out of place

3
h(x): 4

3
4
3 0
4
4 3
4
4 2
3
Greedy Search
 Greedy Search (with Heuristics)
▪ Expand the node that seems closest to the goal

Heuristics h(x)

Cost
178

▪ What can go wrong?

 Greedy 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:
▪ Greedy best-first may take you straight to the (wrong) goal
b
▪ Worst-case: …

▪ Becomes like a badly-guided DFS

▪ Properties:
▪ Exponential both for time and space
▪ Not complete, not optimal
Video of Demo Contours Greedy (Empty)
Video of Demo Contours Greedy (Pacman Small Maze)
 Example: Uniform-Cost Search
State space graph with cost Search Tree

8
e
Cost 1
S 1 a 3 d 2 G
1
c 1 b

Example: Teg Grenager

 Example: Greedy Search
State space graph with heuristics Search Tree

8
e h=1
Cost 1
S 1 a 3 d 2 G
h=6 1 h=5 h=2 h=0
c 1 b
h=7 h=6

Example: Teg Grenager

 Today
▪ Uninformed Search Methods
▪ Depth-First Search
▪ Breadth-First Search
▪ Iterative Deepening Search
▪ Uniform-Cost Search
▪ Informed (Heuristic) Search
▪ Heuristics
▪ Greedy Search
 Next
▪ Informed (Heuristic) Search
▪ Heuristics
▪ Greedy Search
▪ A* Search
▪ Admissibility and optimality
▪ Designing admissible heuristics
▪ Graph search