《人工智能原理人工智能原理 (11).pdf》由会员分享,可在线阅读,更多相关《人工智能原理人工智能原理 (11).pdf(12页珍藏版)》请在得力文库 - 分享文档赚钱的网站上搜索。
1、Principles of Artificial Intelligence1Example ProblemsPrinciples of Artificial Intelligence2 3.2.1 Example1:Vacuum-cleaner world 3.2.2 Example2:8-puzzle 3.2.3 Example3:8-queens problemContentsArtificial Intelligence:Searching:Search3 Vacuum-cleaner world has been introduced in“2.1.6.Intelligent agen
2、t paradigm”.真空吸尘器世界已经在“2.1.6.Intelligent agent paradigm”中讲过。Example 1:Vacuum-cleaner world 真空吸尘器世界3.2.Example Problems The states are determined by both the agent location and dirt location.其状态是由智能体的位置和灰尘的位置决定的。Links denote actions:L=Left,R=Right,S=Suck.链接表示动作:L=左移,R=右移,S=吸尘。Artificial Intelligence:
3、Searching:Search4 States状态 Agent is in one of two locations,each may contain dirt or not.智能体在两个地点中的一个,每个也许有灰尘或者没有。Possible states,2 locations:222=8(n2n).可能的状态,2个地点:222=8(n2n)。1)Initial state 初始状态Any state can be as the initial state.任何状态都可以作为初始状态。2)Actions动作Each state has just three actions:Left,Rig
4、ht,and Suck.每个状态仅有三个动作:左移,右移,以及吸尘。Example 1:Vacuum-cleaner world 真空吸尘器世界3.2.Example ProblemsArtificial Intelligence:Searching:Search5 3)Transition model 转换模型The actions have their expected effects,except that moving:该动作应有的预期效果,下述动作除外:Left in the leftmost,在最左边进行左移 Right in the rightmost,在最右边进行右移 Suck
5、 in a clean square.在清洁区域进行吸尘 4)Goal test目标测试Whether all the squares are clean.是否所有的区域内都干净。5)Path cost路径代价The number of steps in the path(each step costs 1).等于路径的步数(每一步的代价)。Example 1:Vacuum-cleaner world 真空吸尘器世界3.2.Example ProblemsArtificial Intelligence:Searching:Search6 8-puzzle:33 board with 8 num
6、bered tiles and a blank space.8数码难题:33棋盘上有8个数字棋子和一个空格。Example 2:8-puzzle 8数码难题3.2.Example ProblemsA tile adjacent to the blank space can slide into the space.The object is to reach a specified goal state.与空格相邻的滑块可以移向该空格,目的是达到一个指定的目标状态。Start stateGoal stateArtificial Intelligence:Searching:Search7 St
7、ates状态 Each of 8 numbered tiles in one of the 9 squares,and blank in the last square.8个数字滑块每个占据一个方格,而空格则位于最后一个方格。1)Initial state 初始状态 Any state can be the initial state.任意一个状态都可以成为初始状态。2)Actions动作 Simplest formulation defines the actions as movements of the blank space:Left,Right,Up,or Down.最简单的形式化是
8、将动作定义为空格的移动:左、右、上、下。Different subsets are depending on where the blank is.不同的子集依赖于空格的位置。Example 2:8-puzzle8数码难题3.2.Example ProblemsStart stateArtificial Intelligence:Searching:Search8 3)Transition model 转换模型Given a state and action,this returns the resulting state.E.g.,if we apply Left to the start
9、state,the resulting state has the 5 and the blank switched.给定状态和动作,其返回结果状态。例如,如果我们对初始状态施加左移动作,由此产生的状态则使5与空格互换。Example 2:8-puzzle 8数码难题3.2.Example Problems 4)Goal test目标测试Checks whether the state matches the goal configuration.即检查状态是否与目标布局相符。5)Path cost 路经代价The number of steps in the path(each step c
10、osts 1).等于路径的步数(每一步的代价)。TransitionArtificial Intelligence:Searching:Search9 The 8-puzzle belongs to the family of sliding block puzzles,this family is known to be NP-complete.8数码难题属于滑块难题家族,这个家族被认为是NP完的。Sliding block puzzles 滑块难题3.2.Example Problems3x3 sliding puzzle.15-puzzleWord puzzle7x7 sliding b
11、lock puzzle华容道Artificial Intelligence:Searching:Search10A queen attacks another one in the same diagonal.The goal is to place 8 queens on a chessboard such that no queen attacks any other.(A queen attacks any piece in the same row,column or diagonal.)其目标是将8个皇后摆放在国际象棋的棋盘上,使得皇后之间不发生攻击(一个皇后会攻击同一行、同一列或同
12、一斜线上的其他皇后)。Example 3:8-queens problem 8皇后问题3.2.Example Problems Two main kinds of formulation:两种主要类型的形式化方法:Incremental formulation:starts with an empty state,theneach action adds a queen to change the state.增量形式化:从空状态开始,然后每次添加一个皇后改变其状态。Complete-state formulation:starts with all 8 queens on the board
13、,and moves them around.全态形式化:初始时8个皇后都放在棋盘上,然后再将她们移开。Artificial Intelligence:Searching:Search11 States:Any arrangement of 1 to 8 queens on the board is a state.状态:第1至第8个皇后在棋盘上任意摆放,为一个状态。1)Initial state:No queens on the board.初始状态:棋盘上没有皇后。2)Actions:Add a queen to any empty square.动作:添加一个皇后至任意一个空格。3)Tr
14、ansition model:Returns the board with a queen added to the specified square.转换模型:将一个皇后添加到指定空格,再返回该棋局。4)Goal test:8 queens are on the board,none attacked.目标测试:8个皇后都在棋盘上,并且没有攻击。5)Path cost:The number of steps(each step costs 1).路径代价:等于步数(每步代价为1)。Example 3:8-queens problem8皇后问题The incremental formulation 增量式形式化3.2.Example Problems