CHAPTER I
1. definition of Artificial Intelligence
If a machine is able to perform tasks that need "smart",we can consider the machine has a kind of artifical intelligence
Intelligence machine: A kind of machine which can be able to perform various anthropomorphic tasks
independently or interactively under all kinds of environments.
•Artificial Intelligence (discipline) is a branch of computer science involved in research, design and application on intelligent machines. Main objectives are to study how to simulate and execute some intelligent features of human brain using machines, and develop related theory and technologies.（discipline）
•Artificial intelligence (ability) is human intelligence-related functions performed by intelligent machines, i.e.thinking activities, such as judgment,reasoning,evidence,identify,perceive,understanding,design,thinking,planning,learning and problem solving.（ability）
2. GOAL of AI
•Recent goal:
To construct intelligent computer instead of part of human intellectual labor.
•Long-term goal:
To simulate human thinking process and intelligent behavior by automatic machines.
•Ultimate goal:
To achieve various functions of biological intelligence by machine intelligence.
Symbolicism Connectionism Actionism
Symbolism thinks that artificial intelligence comes from mathematical logic. Symbolism is still the mainstream of artificial intelligence. Representatives are Newell, Shaw, Simon and Nilsson and so on.
Connectionism thinks artificial intelligence comes from bionics, especially human brain model.
Intelligence comes from cybernetics. Representative of this school is six-legged walking robot, which is regarded as a new generation of "cybernetic animals", insect behavior control system based on perception - action model.
CHAPTER II 知识表示方法
Fact knowledge
Classification, attributes, relationship, scientific facts, objective facts etc, representing in the form of "... is ...".
Fact is static, as shared, publicly available, proven knowledge, which is a low-level knowledge in knowledge base.
Rule knowledge
A kind of causal knowledge related to behavior and action of problem.
It is dynamic, representing in the form of "if ... then ..." .
Control knowledge
Steps of problem solving
When there are multiple actions, which one should be selected to be active to perform.
e.g. From Beijing to Shanghai, by air or by train?
By air is faster but expensive; by train is slower but cheaper.
Meta knowledge
Meta-knowledge is knowledge about knowledge, which is high-level knowledge in knowledge base, including using rules, interpretation rules, validation rules etc.
e.g. Knowledge catalog
Knowledge Representation
Knowledge representation refers to knowledge symbolization, which can be input into computer.
Representing knowledge according to certain principles and structure.
Explaining meaning of knowledge representation.
Methods
State space method 状态空间
Problem reduction method 问题规约
Predicate logic method 谓词逻辑
Semantic network method 语义网络
Ontology technology ... 本体技术
Frame representation 框架表示
Screenplay representation 剧本表示
Process representation 过程表示
Object-oriented method 面向对象
状态空间法三要点：
状态（state）
State
An ordered set of variables q0, q1 , …, qn describing differences between different things.
Vector representation
Q =[q0, q1,… , qn]T
here, qi (i=0,1，…，n) is component of set, called state variable.
If a set of values of each component are given, a specific state is obtained, such as qk = [q0k, q1k, … ,qnk]T
算符（operator）
Means which makes problem change from one state to another state, called operator.
Operator includes step, procedure, rule, mathematical operator, arithmetic symbol or logical symbol.
状态空间（state space）
Diagram represents all possible states and their relationship of problem.
Including 3 sets,
Set of all possible initial states S
Set of operators F
Set of target states G
State space can be denoted as ternary(S, F, G).
三数码难题和十五数码难题的状态空间表示
状态图示法 state diagram
最短路程
问题归约法 problem reduction
Problem reduction is another method of problem description and problem solving.
Approach
First, divide a problem into sub-problems and sub-sub-problems.
And then solve the minor problem.
Solution to the subset means solution to the original problem.
逆向推理，建立子问题以及子问题的子问题，直到把原始问题规约为一个本原问题
可解节点
二阶汉诺塔
框架表示法 Framework
谓词逻辑表示法 predicate logic
limitations of proposition logic
If we do not quantify the problem, we need to write each separate proposition that people need to study one by one.
语义网络
•Instance relationship
•Generalization relationship
•Assemble relationship
•Attribute relationship
u如果要表示的事实是多元关系，必须将多元关系转化为二元关系。
CHAPTER III 搜索原理
图搜索的基本思想
open list closed list的作用
启发性搜索 informed search
