【人工智能】人工智能导论

什么是人工智能

人工智能：用人工的方法在机器上实现的智能

人工智能学科

智能如何产生

Theory of mind:the core of intelligence is thinking智能通过思维产生—逻辑推理

Knowledge threshold theory:intelligence depends on the amount of knowledge and the degree of generalization知识阈值理论—搜索思想

Evolutionary theory:replacing the representation of knowledge with control进化理论

智能特点

感知能力

Perception:the ability to perceive the external world through the sensory organs

视觉

听觉

记忆和思维能力

逻辑思维logical thinking(abstract thinking)

Like:三段论

Think logical

The thought process is serial

Easy to formalize

Thought process with rigor,reliability

形象思维Image thinking(direct thinking)

Follow your intuition

Thought processes are 

顿悟，灵感Insight thinking(inspired thinking)

Sudden bursts at irregular intervals

学习能力

有指导学习

无指导学习

行为能力

能够表示出来

图灵测试

 

图灵机

 

人工智能的研究内容

知识表示—知识形式化和建模

knowledge representation:the formalization or modeling of human knowledge

符号表示

一阶谓词逻辑

产生式系统

连接机制表示

人工智能网络（深度学习）

机器感知

查看上文【感知能力】

机器思维

通过对信息的处理（机器感知）实现

机器学习

深度学习

Deep learning: It is a hot topic in current research and application.

Deep learning involves building neural networks that mimic the behavior of the human brain. These multi-layered neural networks, like the human brain, can gather information and produce behavior accordingly

SVM：支持向量机

机器行为

研究如何让计算机说，写，画

研究领域

智能信息检索intelligent information retrieval

能理解自然语言

具有推理能力

有一定常识性知识

数据挖掘与知识发现Data mining and knowledge discovery

The purpose of data mining is to find meaningful patterns from a database(a set of rules, clusters, decision trees, knowledge represented by a dependent network, or otherwise).

Data mining processes:

数据预处理

建模

模型评估以及模型应用

专家系统Expert System

Expert systems simulate the thinking process of human experts to solve problems in the field, and their level can reach or even exceed the level of human experts.

自动编程Automatic programming

Program synthesis: The user only needs to tell the computer to "do what", without explaining "how to do", the computer can automatically realize the design of the program.

Program correctness verification: research out a set of theories and methods, through the use of this set of theories and methods can prove the correctness of the program.

机器人Robots

Robot development: program-controlled robot (first generation), adaptive robot (second generation), intelligent robot (modern).

Unimet

Voshatlan

组合优化Combinatorial optimization problem

Np-complete problems：Using the best known methods, the time required to solve the problem increases exponentially with the size of the problem.

Combinatorial optimization problems

人工神经网络Artificial neural network

mp神经网络—感知器—bp神经网络—CNN神经网络

Deep neural network: the current research and application hotspot.

Deep neural network is a multi-level neural network.

Like the human brain, deep neural networks can collect information and generate corresponding behaviors, which can mimic the behavior of the human brain.

These neural networks can gradually perceive and understand the shape and sound of things, that is, "know" things.

分布式人工智能和多智能体系统Distributed Artificial Intelligence and multi-agents

Distributed AI systems take robustness as the standard to control the quality of the system and have interoperability, that is, different heterogeneous systems have the ability to exchange information and work together in a rapidly changing environment.

智能控制Intelligent control

Two salient features of intelligent control are:

First, intelligent control is a control process with a mixture of non-mathematical generalized world model and traditional mathematical model with knowledge representation.

Second, the core of intelligent control is high-level control, its task is to organize the actual environment or process, that is, decision and planning, in order to achieve generalized problem solving.

The basic types of intelligent control are:
(1) expert intelligent control

(2) 模糊控制fuzzy control

(3) neural network control

智能仿真Intelligent simulation

Intelligent simulation is the introduction of AI into the field of simulation and the establishment of intelligent simulation systems.

Simulation is the experiment of dynamic models, that is, the behavior generator drives the model under specified experimental conditions to produce the model behavior. Simulation is to produce conclusive knowledge on the basis of descriptive knowledge, purposive knowledge and processing knowledge.

智能CADIntelligent CAD

Intelligent CAD (ICAD for short) is to introduce artificial intelligence technology into the field of computer-aided design and establish intelligent CAD system. AI can be applied to almost every aspect of CAD technology. From the current development trend, there are at least the following four aspects:

(1)设计自动化 Design automation.

(2) Intelligent interaction.

(3) Intelligent graphics.

(4) Automatic data acquisition

智能CAI Intelligent CAI

Intelligent CAI is to introduce AI into the field of computer-aided instruction. ICAI systems are generally divided into specialized knowledge, teaching strategies, and intelligent interfacesfor student models and natural language.

ICAI should have the following intelligence characteristics:

(1) Automatically generate questions and exercises.

(2) Automatically select and adjust the teaching content and progress according to the

students' learning situation.

(3) Automatically solve problems and generate answers on the basis of understanding the

teaching content.

(4) Ability to generate and understand naturallanguage.

(5) Have the ability to understand and consult the teaching content.

(6) Able to diagnose students' mistakes, analyze the causes and take corrective measures.

(7) Able to evaluate students' learning behavior.

(8) Can continuously improve teaching strategiesin teaching.

智能管理和智能决策Intelligent management and intelligent decision

Intelligent management is to introduce artificial intelligence technology into the field of management, establish intelligent management system, study how to improve the intelligence level of computer management system, as well as the design theory, method and implementation technology of intelligent management system.

Intelligent decision is the introduction of artificial intelligence technology into the decision-making process,the establishment of intelligent decision supportsystem.

智能多媒体Intelligent multimedia system

Multimedia computer system is a computer system that can comprehensively process various media information such astext, graphics, images and sound.

Intelligent multimedia is to introduce artificial intelligence technology into multimedia system,so that its function and performance can be further developed and improved.

智能操作系统Intelligent operating system

 

Intelligent computer system

 

智能通信Intelligent Communication

 

智能网络系统Intelligent network system

 

人工生命Artificial Life

 

知识表示knowledge representation

什么是知识

知识的特征

不确定性uncertainty

可表示性和可应用性

知识分类

scope

常识

专业知识

按功能性分类

事实性知识

控制性知识

过程性知识

知识表示

人类知识的形式化与模型化

一组转换规则的集合

 

状态空间表示State space

状态：n元组

操作符（算子）

1.设定初始状态

2.设定目标状态

3.经过一系列操作后，将初始状态转变为目标状态

15数码问题

在4x4的棋盘里放入15个数字，一个数码排列方式代表一个状态

操作最多有四个，但避免还原，需要去掉一个操作

状态图：有向图，状态用节点来表示，有向边表示操作

问题规约表示法

把问题分解为原子性的子问题

规约图

一阶谓词逻辑

 

产生式系统

框架语意网络