Search

Agents

Uninformed
search
[Some slides created by Dan Klein and Pieter Abbeel for Intro to AI at UC Berkeley (ai.berkeley.edu). Some slides were taken from Columbia University AI course]
https://st3.depositphotos.com/1005404/18747/i/1600/depositphotos_187473334-stock-photo-search-white-robot-and-lens.jpg
Uninformed
search
- Use no domain knowledge!

- Uninformed search strategies do not

take into account the location of the
goal.

- Intuitively, these algorithms ignore

where they are going until they find a
goal and report success.

Uninformed agents searching for the goal.

Uninformed search
Use no domain knowledge!
Strategies:
1. Breadth-first search (BFS): Expand shallowest node
2. Depth-first search (DFS): Expand deepest node
3. Depth-limited search (DLS): Depth first with depth limit
4. Iterative-deepening search (IDS): DLS with increasing limit
5. Uniform-cost search (UCS): Expand least cost node
Breadth-first
search (BFS)
BFS: Expand shallowest first.
BFS Search
Pseudo-code
 BFS
criteria?

This is a gif, wait for it to start 

Breadth-first search (BFS)
• Complete: Yes (if b is finite)
• Time: 1 + b + + + . . . + = O()
• Space: O()
◦ Note: If the goal test is applied at expansion rather than generation then O()

• Optimal: Yes (if cost = 1 per step).

• Implementation: fringe: FIFO (Queue)
How bad is BFS?

Time and Memory

requirements for breadth-
first search for a branching
factor b=10; 1 million
nodes per second; 1,000
bytes per node.
Memory requirement &
exponential time
complexity are the biggest
handicaps of BFS!
Question: If time and space complexities are exponential,
why use BFS?
Depth-first
search (DFS)
DFS: Expand deepest first.
DFS Search
Pseudo-code.
 DFS
criteria?
This is a gif, wait for it to start 
Depth-first search (DFS)
• Complete: No, fails in infinite-depth spaces & spaces with loops
◦ Modify to avoid repeated states along path.
◦ ⇒ complete in finite spaces

• Time: O(): 1 + b + + + . . . + = O()

◦ Bad if m is much larger than d but if solutions are dense may be much faster than BFS.

• Space: O(bm) linear space complexity!

◦ (needs to store only a single path from the root to a leaf node, along with the remaining
unexpanded sibling nodes for each node on the path, hence the m factor.)
• Optimal: No
• Implementation: fringe: LIFO (Stack)
How bad is DFS?
For depth = 16 we go
down from 10 exabytes
in BFS to ??? in DFS.

156 kilobytes!
Quiz: DFS vs BFS
When will BFS outperform DFS?

When will DFS outperform BFS?

[Demo: dfs/bfs maze water (L2D6)]

Video of Demo Maze Water DFS/BFS
(part 1)
Video of Demo Maze Water DFS/BFS
(part 2)
Depth-limited search
• DFS with depth limit l (nodes at level l has no successors).
• Select some limit L in depth to explore with DFS

https://slideplayer.com/slide/12719512/
If have some knowledge about the
problem, may be we don’t need to
go to a full depth.

For example, if we know that any

city can be reached from another
city in at most L steps, with L < 36,
we can stop expanding at 36th step.
Depth-limited
search
Pseudo-code.
Iterative-deepening search
• Combines the benefits of BFS and DFS.
• Idea: Iteratively increase the search limit until the depth of the shallowest
solution d is reached.
• Applies DLS with increasing limits.
• The algorithm will stop if a solution is found or if DLS returns a failure (no
solution).
• Because most of the nodes are on the bottom of the search tree, it not a big
waste to iteratively re-generate the top
• Let’s take an example with a depth limit between 0 and 3.
 Iterative Deepening
Idea: get DFS’s space advantage with BFS’s time
/ shallow-solution advantages
b
◦ 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. …..

Isn’t that wastefully redundant?

◦ Generally most work happens in the lowest level
searched, so not so bad!
Iterative Deepening
Limit = 0
Depth-
limited
search
(Limit =
1)
Depth-
limited
search
(Limit = 2)
Depth-
limited
search
(Limit = 3)
Uniform-cost
search
• The arcs in the search
graph may have weights
(different cost attached).
How to leverage this
information?

https://ds055uzetaobb.cloudfront.net/image_optimizer/5f3b305766eb8a8847f54619c1e697f36cd7b1e6.png
For this
problem
• BFS will find the shortest
path which may be costly.
• What if we want the
cheapest not shallowest
solution?
• Modify BFS: Prioritize by
cost not depth → Expand
node n with the lowest path
cost g(n)
• Explores increasing costs.
Uniformed-
cost search
Pseudo-code
Uniform-cost search
Let’s say we need to go from Baku to
Sheki. 70
Sheki Ganja

BFS would find route: Baku -> Sumgayit ->

230
Sheki, because it need to expand only

350
twice. Total distance = 30 + 350 = 380

UCS would find route: Sumgayit Xirdalan

Baku ->Xirdalan -> Ganja -> Sheki,
30
because it expands the least-cost node 10
first. Total distance = 10 + 230 + 70 = 310, Baku
which is lower than BFS route.
Uniform-cost search
• Complete: Yes Assuming best solution has a finite cost and minimum arc cost is positive.
• Time: O()
◦ Suppose C∗: cost of the optimal solution
◦ Every action costs at least ϵ (bound on the cost)
◦ The effective depth is roughly C∗/ϵ (how deep the cheapest solution could be).

• Space: O() (exponential in effective depth)

• Optimal: Yes
• Implementation: fringe = queue ordered by path cost g(n)
Uniform-cost
search
Strategy: expand lowest
path cost
The good: UCS is complete
and optimal!
The bad:
- Explores options in every
“direction”
- No information about goal
location
 Uniform Cost Search
2 a G
Strategy: expand a cheapest b c 3
node first: 1 8 2
2 e
3 d f
Fringe is a priority queue 9 2
S h 8
(priority: cumulative cost) 1
1 p q r
15
 Uniform Cost Search
2 a G
Strategy: expand a cheapest b c 3
node first: 1 8 2
2 e
3 d f
Fringe is a priority queue 9 2
S h 8
(priority: 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
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?
(part 1)
Video of Demo Maze with Deep/Shallow Water --- DFS, BFS, or UCS? (part 2)
Video of Demo Maze with Deep/Shallow Water --- DFS, BFS, or UCS? (part 3)
Examples using the map
Start: Las Vegas
Goal: Calgary

BFS

Order of Visit: Las Vegas, Los Angeles, Salt Lake City, El Paso, Phoenix, San Francisco,
Denver, Helena, Portland, Dallas, Santa Fe, Kansas City, Omaha, Calgary.
Examples using the map
Start: Las Vegas
Goal: Calgary

DFS

Order of Visit: Las Vegas, Los Angeles, El Paso, Dallas, Houston, New Orleans, Atlanta,
Charleston, Nashville, Saint Louis, Chicago, Duluth, Helena, Calgary.
Examples using the map
Start: Las Vegas
Goal: Calgary

UCS

Order of Visit: Las Vegas, Los Angeles, Salt Lake City, San Francisco, Phoenix, Denver, Helena,
El Paso, Santa Fe, Portland, Seattle, Omaha, Kansas City, Calgary.
 8 Puzzle
https://kartikkukreja.files.wordpress.com/2015/06/8puzzle1.jpg
Example DFS

https://www.darpa.mil/ 43
Example
BFS

https://www.darpa.mil/ 44
Summary of algorithms