A General Introduction to Artificial Intelligence

Outline Best-first search. Best-first search. Greedy best-first search Greedy best-first search A * search. A * search. Characteristics of Heuristics Characteristics of Heuristics Variations of A* Variations of A* Memory Bounded A*. Memory Bounded A*. Online (Real Time) A*. Online (Real Time) A*. Local search algorithms Local search algorithms Hill-climbing search. Hill-climbing search. Simulated annealing search. Simulated annealing search. Gradient descent search. Gradient descent search.

Review: Tree Search A search strategy is defined by picking the order of node expansion A search strategy is defined by picking the order of node expansion

Best-First Search Idea: use an evaluation function f(n) for each node Idea: use an evaluation function f(n) for each node estimate of "desirability" estimate of "desirability" Expand most desirable unexpanded node. Expand most desirable unexpanded node. Implementation: Implementation: Order the nodes in fringe in decreasing order of desirability. Special cases: Special cases: greedy best-first search greedy best-first search A * search. A * search.

Best-First Search

Romania with step costs in km Romania with step costs in km

Greedy Best-First Search

Evaluation function f(n) = h(n) (heuristic) Evaluation function f(n) = h(n) (heuristic) = estimate of cost from n to goal. = estimate of cost from n to goal. e.g., h SLD (n) = straight-line distance from n to Bucharest. e.g., h SLD (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 expands the node that appears to be closest to goal.

Greedy Best-First Search example

Properties of Greedy BFS Complete? No – can get stuck in loops, e.g., Iasi  Neamt  Iasi  Neamt  Complete? No – can get stuck in loops, e.g., Iasi  Neamt  Iasi  Neamt  Time? O(b m ), but a good heuristic can give dramatic improvement. Time? O(b m ), but a good heuristic can give dramatic improvement. Space? O(b m ) -- keeps all nodes in memory. Space? O(b m ) -- keeps all nodes in memory. Optimal? No. Optimal? No.

A * Search Idea: avoid expanding paths that are already expensive. Idea: avoid expanding paths that are already expensive. Evaluation function f(n) = g(n) + h(n). Evaluation function f(n) = g(n) + h(n). g(n) = cost so far to reach n g(n) = cost so far to reach n h(n) = estimated cost from n to goal h(n) = estimated cost from n to goal f(n) = estimated total cost of path through n to goal. f(n) = estimated total cost of path through n to goal.

A * Search

A * Search Example

Admissible Heuristics A heuristic h(n) is admissible if for every node n, 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 An admissible heuristic never overestimates the cost to reach the goal, i.e., it is optimistic Example: h SLD (n) (never overestimates the actual road distance) Example: h SLD (n) (never overestimates the actual road distance) Theorem: If h(n) is admissible, A * using TREE- SEARCH is optimal. Theorem: If h(n) is admissible, A * using TREE- SEARCH is optimal.

Optimality of A * (proof) Suppose some suboptimal goal G 2 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. Suppose some suboptimal goal G 2 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(G 2 ) = g(G 2 )since h(G 2 ) = 0 f(G 2 ) = g(G 2 )since h(G 2 ) = 0 g(G 2 ) > g(G) since G 2 is suboptimal g(G 2 ) > g(G) since G 2 is suboptimal f(G) = g(G)since h(G) = 0 f(G) = g(G)since h(G) = 0 f(G 2 ) > f(G)from above f(G 2 ) > f(G)from above

Optimality of A * (proof) Suppose some suboptimal goal G 2 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. Suppose some suboptimal goal G 2 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(G 2 )> f(G) from above f(G 2 )> f(G) from above h(n)≤ h^*(n)since h is admissible h(n)≤ h^*(n)since h is admissible g(n) + h(n)≤ g(n) + h * (n) g(n) + h(n)≤ g(n) + h * (n) f(n) ≤ f(G) f(n) ≤ f(G) Hence f(G 2 ) > f(n), and A * will never select G 2 for expansion.

Consistent Heuristics A heuristic is consistent if for every node n, every successor n' of n generated by any action a, 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 If h is consistent, we have f(n') = g(n') + h(n') = g(n) + c(n,a,n') + h(n') = g(n) + c(n,a,n') + h(n') ≥ g(n) + h(n) ≥ g(n) + h(n) = f(n) = f(n) i.e., f(n) is non-decreasing along any path. i.e., f(n) is non-decreasing along any path. Theorem: If h(n) is consistent, A* using GRAPH-SEARCH is optimal Theorem: If h(n) is consistent, A* using GRAPH-SEARCH is optimal

Optimality of A * A * expands nodes in order of increasing f value A * expands nodes in order of increasing f value Gradually adds "f-contours" of nodes Gradually adds "f-contours" of nodes Contour i has all nodes with f=f i, where f i < f i+1 Contour i has all nodes with f=f i, where f i < f i+1

Properties of A* Complete? Yes (unless there are infinitely many nodes with f ≤ f(G) ) Complete? Yes (unless there are infinitely many nodes with f ≤ f(G) ) Time? Exponential Time? Exponential Space? Keeps all nodes in memory Space? Keeps all nodes in memory Optimal? Yes Optimal? Yes

Admissible Heuristics E.g., for the 8-puzzle: h 1 (n) = number of misplaced tiles h 1 (n) = number of misplaced tiles h 2 (n) = total Manhattan distance h 2 (n) = total Manhattan distance (i.e., no. of squares from desired location of each tile) h 1 (S) = ? h 1 (S) = ? h 2 (S) = ? h 2 (S) = ?

Admissible Heuristics E.g., for the 8-puzzle: h 1 (n) = number of misplaced tiles h 1 (n) = number of misplaced tiles h 2 (n) = total Manhattan distance h 2 (n) = total Manhattan distance (i.e., no. of squares from desired location of each tile) h 1 (S) = ? 8 h 1 (S) = ? 8 h 2 (S) = ? = 18 h 2 (S) = ? = 18

Dominance If h 2 (n) ≥ h 1 (n) for all n (both admissible) If h 2 (n) ≥ h 1 (n) for all n (both admissible) then h 2 dominates h 1 then h 2 dominates h 1 h 2 is better for search h 2 is better for search Typical search costs (average number of nodes expanded): Typical search costs (average number of nodes expanded): d=12IDS = 3,644,035 nodes A * (h 1 ) = 227 nodes A * (h 2 ) = 73 nodes d=12IDS = 3,644,035 nodes A * (h 1 ) = 227 nodes A * (h 2 ) = 73 nodes d=24 IDS = too many nodes A * (h 1 ) = 39,135 nodes A * (h 2 ) = 1,641 nodes d=24 IDS = too many nodes A * (h 1 ) = 39,135 nodes A * (h 2 ) = 1,641 nodes

Relaxed Problems A problem with fewer restrictions on the actions is called a relaxed problem 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 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 h 1 (n) gives the shortest solution If the rules of the 8-puzzle are relaxed so that a tile can move anywhere, then h 1 (n) gives the shortest solution If the rules are relaxed so that a tile can move to any adjacent square, then h 2 (n) gives the shortest solution If the rules are relaxed so that a tile can move to any adjacent square, then h 2 (n) gives the shortest solution

Variations of A* Limitation of A* is memory consuming. Limitation of A* is memory consuming. The main idea is to mimic depth-first search, iterative deepening search (using linear space and branch and bound technique). The main idea is to mimic depth-first search, iterative deepening search (using linear space and branch and bound technique). Iterative Deepening A* (IDA): use f=g+h as the cutoff value. More recent improvements: Recursive Best First Search (RBFS) and Memory- Bounded A* (MA*), Simplified Memory Bounded A*. Iterative Deepening A* (IDA): use f=g+h as the cutoff value. More recent improvements: Recursive Best First Search (RBFS) and Memory- Bounded A* (MA*), Simplified Memory Bounded A*.

Iterated Deepening A*

Recursive Best First Search

Online Search Algorithms In many problem the requirement for search is online and real time. For such problems, it can only be solved by agent action executions. In many problem the requirement for search is online and real time. For such problems, it can only be solved by agent action executions. A online search agent just know the following: A online search agent just know the following: Actions(s) - a list of possible action at node s. Actions(s) - a list of possible action at node s. C(s,a,s') - Step cost function. Cannot be used until agent know that s' is an outcome. C(s,a,s') - Step cost function. Cannot be used until agent know that s' is an outcome. Goal-Test(s) - to test is s is a goal Goal-Test(s) - to test is s is a goal Performance measure: Compression Ratio -the ratio of cost of actual path taken by agent and the optimal path if the agent know the search space beforehand Performance measure: Compression Ratio -the ratio of cost of actual path taken by agent and the optimal path if the agent know the search space beforehand

Online Search Algorithms

Local Search Algorithms In many optimization problems, the path to the goal is irrelevant; the goal state itself is the solution In many optimization problems, the path to the goal is irrelevant; the goal state itself is the solution State space = set of "complete" configurations State space = set of "complete" configurations Find configuration satisfying constraints, e.g., n- queens Find configuration satisfying constraints, e.g., n- queens In such cases, we can use local search algorithms In such cases, we can use local search algorithms keep a single "current" state, try to improve it keep a single "current" state, try to improve it

Example: n-queens Put n queens on an n × n board with no two queens on the same row, column, or diagonal. Put n queens on an n × n board with no two queens on the same row, column, or diagonal.

Hill-Climbing Search

Problem: depending on initial state, can get stuck in local maxima. Problem: depending on initial state, can get stuck in local maxima.

Example: 8-queens problem h = number of pairs of queens that are attacking each other, either directly or indirectly h = number of pairs of queens that are attacking each other, either directly or indirectly h = 17 for the above state h = 17 for the above state

Example: 8-queens problem A local minimum with h = 1 A local minimum with h = 1

Local Search in Continuous Domains - Gradient Descent p k - search direction  k - search step size (learning factor) Or

Gradient Descent Algorithm Choose the next step so that: With small step size we can approximate F(x): Where To make the function value decrease: Steepest Descent Direction:

Example

Convergence of Algorithm

Simulated Annealing Search Idea: escape local maxima by allowing some "bad" moves but gradually decrease their frequency Idea: escape local maxima by allowing some "bad" moves but gradually decrease their frequency

Properties of SA Search One can prove: If T decreases slowly enough, then simulated annealing search will find a global optimum with probability approaching 1. One can prove: If T decreases slowly enough, then simulated annealing search will find a global optimum with probability approaching 1. Widely used in VLSI layout, airline scheduling, etc. Widely used in VLSI layout, airline scheduling, etc.

Local Beam Search Keep track of k states rather than just one Keep track of k states rather than just one Start with k randomly generated states Start with k randomly generated states At each iteration, all the successors of all k states are generated 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. If any one is a goal state, stop; else select the k best successors from the complete list and repeat.

Further Reading Text book : chapter 4. Text book : chapter 4. J. Pearl, Heuristics, Addison-Wesley Publishing, J. Pearl, Heuristics, Addison-Wesley Publishing, E. Aarts and J. Korst, Simulated Annealing and Boltzman Machines, Addison-Wesley Publishing, E. Aarts and J. Korst, Simulated Annealing and Boltzman Machines, Addison-Wesley Publishing, E. Aarts and J. Lenstra, Local Search in Combinatorial Optimization, Princeton University Press, E. Aarts and J. Lenstra, Local Search in Combinatorial Optimization, Princeton University Press, 2003.