1 CSE 4705 Artificial Intelligence Jinbo Bi Department of Computer Science & Engineering
2 TodayToday Intelligent Agents
3 Inverted pendulum Example to demonstrate a learning agent
4 8-puzzle8-puzzle A tile adjacent to the blank space can slide into the space.
5 Holiday in Romania Start Goal
6 Complexity of Breadth-First Search
7 Holiday in Romania Start Goal
8 ComparisonComparison
9 Demonstration on Games/Robots Breadth First Search Pink: starting point Blue: goal Teal: scanned squares Darker: closer to starting point
10 Demonstration on Games/Robots An optimal informed search algorithm A* We add a heuristic estimate of distance to the goal Yellow: examined nodes with high h(n) Blue: examined nodes with low h(n)
11 Demonstration on Games/Robots Breadth-first search expands many many nodes Pink: starting node Dark blue: goal
12 Demonstration on Games/Robots A* search expands much fewer nodes Pink: starting node Dark blue: goal
13 Start Goal The distance from each city to Bucharest:
14 Best-first Search
15 Best-first Search
16 A* Search
17 A* Search
18 A* Search
19 Hill Climbing
20 8-puzzle8-puzzle Start Goal
21 Hill-Climbing Ex: 8-queens
22 Gradient ascent/descent
23 Gradient methods / Newton’s methods Contour lines of a function (Green: gradient descent, Red: Newton’s methods)
24 Difficult Problems
25 Difficult Problems
26 Random Restart
27 Genetic Algorithm A short video explains Genetic Algorithm in 3 minutes
28 Genetic Algorithm
29 Searching nondeterministic The 8 physical states of the vacuum world
30 Searching nondeterministic Fig. 4.10, AND-OR Search Tree, and a depth-first search
31 Searching nondeterministic Fig. 4.11, AND-OR Search algorithm (graph search) and a depth-first search, it returns a conditional plan that reaches a goal state in all circumstances S i in
32 Searching partial observable Deterministic Non-deterministic Fig. 4.13
33 Searching partial observable
34 Searching partial observable A vacuum has local sensors, and can report a state of [location, dirty/clean]
35 Searching partial observable Partial observations can still be quite useful (Fig. 4.18
36 Game Tree for Tic-Tac-Toe
37 An Evaluation Function for Tic-Tac-Toe f(n) = 8-8=0 f(n) = 8-5=3 f(n) = 8-6=2 f(n) = 2f(n) = 3 f(n): the potential # of lines with 3 x – the potential # of lines with three o f(n) = 0 if a tie f(n) = + ∞ if n is a terminal win f(n) = - ∞ if n is a terminal loss
38 Two Players MINIMAX value for a Two-Play Game Tree
39 Multiple Players
40 Alpha-Beta Pruning
41 Map Coloring
42 A Consistent and Complete Solution to Map Coloring
43 BacktrackingBacktracking
44 Backtracking – Map Coloring
45 Improving Backtracking Most constrained variables Most constraining variables
46 Improving Backtracking Given n variables, choose the least constraining value
47 Improving Backtracking Forward checking
48 Arc Consistency
49 ≠ General Backtracking
50 The Wumpus World Flash-Game.htm