The document discusses different types of knowledge representations that can be used for search problems:
- Atomic representation treats states as indivisible units without internal structure. Search operates directly on atomic states.
- Propositional/factored representation describes states using state variables that can take on different values. Search and inference is done in logical and probabilistic models over the variables.
- Relational representation describes states using objects and relations between objects. Search and inference is done in predicate logic or relational probabilistic models.
The document uses the example of a vacuum world problem to illustrate the different representations and how the state space changes based on the number of rooms and robots. It then discusses different uninformed search strategies
The document discusses different types of knowledge representations that can be used for search problems:
- Atomic representation treats states as indivisible units without internal structure. Search operates directly on atomic states.
- Propositional/factored representation describes states using state variables that can take on different values. Search and inference is done in logical and probabilistic models over the variables.
- Relational representation describes states using objects and relations between objects. Search and inference is done in predicate logic or relational probabilistic models.
The document uses the example of a vacuum world problem to illustrate the different representations and how the state space changes based on the number of rooms and robots. It then discusses different uninformed search strategies
The document discusses different types of knowledge representations that can be used for search problems:
- Atomic representation treats states as indivisible units without internal structure. Search operates directly on atomic states.
- Propositional/factored representation describes states using state variables that can take on different values. Search and inference is done in logical and probabilistic models over the variables.
- Relational representation describes states using objects and relations between objects. Search and inference is done in predicate logic or relational probabilistic models.
The document uses the example of a vacuum world problem to illustrate the different representations and how the state space changes based on the number of rooms and robots. It then discusses different uninformed search strategies
The document discusses different types of knowledge representations that can be used for search problems:
- Atomic representation treats states as indivisible units without internal structure. Search operates directly on atomic states.
- Propositional/factored representation describes states using state variables that can take on different values. Search and inference is done in logical and probabilistic models over the variables.
- Relational representation describes states using objects and relations between objects. Search and inference is done in predicate logic or relational probabilistic models.
The document uses the example of a vacuum world problem to illustrate the different representations and how the state space changes based on the number of rooms and robots. It then discusses different uninformed search strategies
Chapter 3 (Based on slides by Stuart Russell, Subbarao Kambhampati, Dan Weld, Oren Etzioni, Henry Kautz, Richard Korf, and other UW-AI faculty) Agent’s Knowledge Representation
Type State representation Focus
Atomic States are indivisible; Search on atomic states; No internal structure Propositional States are made of state Search+inference in (aka Factored) variables that take values logical (prop logic) and (Propositional or Multi- probabilistic (bayes nets) valued or Continuous) representations Relational States describe the Search+Inference in objects in the world and predicate logic (or their inter-relations relational prob. Models) First-order +functions over objects Search+Inference in first order logic (or first order probabilistic models) Illustration with Vacuum World Propositional/Factored: States made up of 3 state variables Atomic: Dirt-in-left-room T/F Dirt-in-right-room T/F S1, S2…. S8 Roomba-in-room L/R state is seen as an indivisible Each state is an assignment of snapshot Values to state variables 23 Different states All Actions are SXS matrices.. Actions can just mention the variables they affect If you add a second roomba the state space doubles Note that the representation is compact (logarithmic in the Relational: size of the state space) If you want to consider noisiness World made of objects: Roomba; L-room, R-room Of the rooms, the representation Relations: In (<robot>, <room>); dirty(<room>) If you add a second roomba, the Quadruples.. Representation increases by just one If you add a second roomba, or more rooms, only the objects increase. More state variable. If you want to consider “noisiness” of If you want to consider noisiness, you just need to add one other relation rooms, we need two variables, one for Atomic Agent Input: – Set of states – Operators [and costs] – Start state – Goal state [test]
Output: • Path: start a state satisfying goal test • [May require shortest path]
4 What is Search? • Search is a class of techniques for systematically finding or constructing solutions to problems.
• Example technique: generate-and-test.
• Example problem: Combination lock.
1. Generate a possible solution.
2. Test the solution. 3. If solution found THEN done ELSE return to step 1. 5 Why is search interesting? • Many (all?) AI problems can be formulated as search problems!
• Examples: • Path planning • Games • Natural Language Processing • Machine learning • …
6 Example: The 8-puzzle
• states? • actions? • goal test? • path cost?
7 Example: The 8-puzzle
• states? locations of tiles
• actions? move blank left, right, up, down • goal test? = goal state (given) • path cost? 1 per move •
• [Note: optimal solution of n-Puzzle family is NP-hard]
8 Search Tree Example: Fragment of 8-Puzzle Problem Space
9 Example: robotic assembly
• states?: real-valued coordinates of robot joint angles parts of the object to be
assembled • • actions?: continuous motions of robot joints • • goal test?: complete assembly • • path cost?: time to execute • 10 Example: Romania • On holiday in Romania; currently in Arad. • Flight leaves tomorrow from Bucharest • • Formulate goal: – be in Bucharest – • Formulate problem: – states: various cities – actions: drive between cities – • Find solution: – sequence of cities, e.g., Arad, Sibiu, Fagaras, Bucharest –
11 Example: N Queens • Input: Q – Set of states Q Q – Operators [and costs] Q
– Start state
– Goal state (test)
• Output
12 Implementation: states vs. nodes
• A state is a (representation of) a physical configuration
• A node is a data structure constituting part of a search tree includes state, parent node, action, path cost g(x), depth
• The Expand function creates new nodes, filling in the various fields and using the SuccessorFn of the problem to create the corresponding states. •
13 Search strategies • A search strategy is defined by picking the order of node expansion • Strategies are evaluated along the following dimensions: – completeness: does it always find a solution if one exists? – time complexity: number of nodes generated – space complexity: maximum number of nodes in memory – optimality: does it always find a least-cost solution? – systematicity: does it visit each state at most once?
• Time and space complexity are measured in terms of
– b: maximum branching factor of the search tree – d: depth of the least-cost solution – m: maximum depth of the state space (may be ∞)
14 Uninformed search strategies
• Uninformed search strategies use only the information
available in the problem definition
• Breadth-first search
• Depth-first search
• Depth-limited search
• Iterative deepening search
15 Repeated states • Failure to detect repeated states can turn a linear problem into an exponential one! •
16 Depth First Search • Maintain stack of nodes to visit • Evaluation – Complete? No
– Time Complexity? O(b^m) a – Space Complexity? O(bm) b c
d e f g h
http://www.youtube.com/watch?v=dtoFAvtVE4U 17 Breadth First Search: shortest first • Maintain queue of nodes to visit • Evaluation – Complete? Yes (b is finite) – Time Complexity? O(b^d) a – Space Complexity? O(b^d) b c – Optimal? Yes, if stepcost=1 d e f g h
18 Uniform Cost Search: cheapest first • Maintain queue of nodes to visit • Evaluation – Complete? Yes (b is finite) – Time Complexity? O(b^(C*/e)) a 1 5 – Space Complexity? O(b^(C*/e)) b c – Optimal? Yes 2 6 1 3 4
• Optimal? – Yes, if step cost = 1 – Can be modified to explore uniform cost tree (iterative lengthening)
• Systematic?
28 Cost of Iterative Deepening b ratio ID to DLS
2 3
3 2
5 1.5
10 1.2
25 1.08
100 1.02
29 Summary of algorithms
31 A BFS: A,B,G DFS: A,B,C,D,G B IDDFS: (A), (A, B, G)
C Note that IDDFS can do fewer
D expansions than DFS on a graph
G shaped search space. BFS: A,B,G A
DFS: A,B,A,B,A,B,A,B,A,B B IDDFS: (A), (A, B, G)
C Note that IDDFS can do fewer
D expansions than DFS on a graph
G shaped search space.
Search on undirected graphs or directed graphs with cycles…
Cycles galore… Graph (instead of tree) Search: Handling repeated nodes • Repeated expansions is a bigger issue for DFS than for BFS or IDDFS • Trying to remember all previously expanded nodes and comparing the new nodes with them is infeasible • Space becomes exponential • duplicate checking can also be expensive
• Partial reduction in repeated expansion can be done by
• Checking to see if any children of a node n have the same state as the parent of n • Checking to see if any children of a node n have the same state as any ancestor of n (at most d ancestors for n—where d is the depth of n) Forwards vs. Backwards
35 vs. Bidirectional
When is bidirectional search applicable?
• Generating predecessors is easy • Only 1 (or few) goal states 36 Bidirectional search • Complete? Yes
• Time? – O(bd/2)
• Space? – O(bd/2)
• Optimal? – Yes if uniform cost search used in both directions 37 Breadth-First goes level by level Visualizing Breadth-First & Uniform Cost Search This is also a proof of optimality…
