Chapter One Introduction
1. Definition of artificial intelligence
Definition : A software or program that enables a machine to learn from data through certain algorithms and use what it learns to make human-like decisions.
2. Various cognitive views of artificial intelligence
- The principle of symbolism : based on the assumption of a physical symbol system and the principle of limited rationality, it originated from mathematical logic ;
- The principle of connectionism : based on the neural network and the connection mechanism and learning algorithm between them, it originated from bionics , especially the research on the human brain model;
- The principle of actionism : based on cybernetics and perception-action control system, originated from cybernetics .
3. Elements of artificial intelligence
- Knowledge : the objective laws of the world that people know through experience, learning or association
- Data ( Data ) : the results of facts or observations, referring to all the numbers, letters, symbols, image signals and analog quantities that can be input into the computer and processed by the program.
- Algorithm : An accurate and complete description of the problem-solving scheme, which is a series of clear instructions for solving the problem, and represents a systematic method to describe the strategy mechanism of problem- solving .
- Computing power : The mathematical induction and transformation ability of a machine, that is, the ability to convert abstract and complex mathematical expressions or numbers into understandable mathematical formulas through mathematical methods.
- Talent _ _ _
4. Classification of artificial intelligence systems
expert system ( expert system )
The fuzzy logic system ( The fuzzy logic system )
Neural network system
Machine learning system ( Machine learning system )
Bionic evolutionary system
Swarm intelligence system
Distributed intelligent system
Integrated intelligent system
Autonomous intelligent system
Man-machine coordinated intelligence system
5. Basic content of artificial intelligence research
Cognitive Modeling : Halston summarizes cognition into five models
Knowledge representation : state space, problem reduction, predicate logic
Knowledge reasoning : deductive reasoning, inductive reasoning, analogical reasoning
Computational Intelligence : Neural Computing, Fuzzy Computing, Evolutionary Computing
Knowledge application : expert system, machine learning, automatic planning
Machine perception : pattern recognition, natural language processing
Machine thinking : comprehensive knowledge representation, knowledge reasoning, etc.
Machine Learning : Machines acquire knowledge and learn automatically
Machine Behavior : Expressiveness and Action
Intelligent system construction : distributed system, parallel processing system
6. What are the main research and application areas of artificial intelligence? Among them, which are new research hotspots?
The main application fields of artificial intelligence are: fingerprint recognition, retinal recognition, face recognition, machine vision, expert system.
The main research of artificial intelligence includes: cognitive science, machine learning, natural language processing, robotics, computer game,
Automatic theorem proving, pattern recognition, computer vision, artificial neural network, expert system, knowledge discovery and data mining, automatic programming, intelligent control, intelligent decision support system, distributed artificial intelligence.
Research Hotspot: Deep Learning
7. Artificial intelligence research methods
Functional Simulation
Structural simulation method
Behavioral Simulation
integrated simulation
Chapter 2 - Knowledge Representation Methods
1. Composition of the problem specification
- an initial problem description;
- A set of operators that transform a problem into subproblems;
- A set of source problem descriptions.
2. Missionary Questions (Completed)
3. Shortest travel problem (completed)
Chapter 3 - Search Reasoning Techniques
1. Breadth-first search algorithm :
- Put the start node in the OPEN list (if the start node is a goal node, then find a solution).
- If OPEN is an empty list, there is no solution, and exit fails; otherwise, continue.
- Remove the first node (node n) from the OPEN list and put it in the extended node list of CLOSED.
- Expand node n. If there is no successor node, go to step 2 above
- Put all successor nodes of n at the end of the OPEN list, and provide pointers from these successor nodes back to n.
- If any successor node of n is a target node, find a solution and exit successfully; otherwise, go to step 2.
Algorithm block diagram:
2. Depth-first search algorithm :
- Put the starting node S into the OPEN list of unexpanded nodes. If the node is a goal node, a solution is obtained.
- If OPEN is an empty list, fail and exit.
- Move the first node (node n) from the OPEN list to the CLOSED list.
- If the depth of node n is equal to the maximum depth, go to 2.
- Expand node n, generate all its descendants, and put them at the head of the OPEN list. If there are no descendants, turn to 2.
- If any of the successor nodes is the target node, a solution is obtained and exit is successful; otherwise, turn to 2. Obviously, the order in which nodes enter and exit the OPEN table in the depth-first algorithm is last-in-first-out, and the OPEN table is a stack
Algorithm block diagram:
3. Breadth-first search method to solve the eight-number puzzle (completed)
4. Bounded depth-first search method to solve the eight-digit puzzle (completed)
5. Probabilistic reasoning method calculation question 1 (one conclusion E supports multiple hypotheses H1, H2, H3) (completed)
6. Probabilistic reasoning method calculation question 2 (multiple conclusions E1, E2 support multiple hypotheses H1, H2, H3) (completed)
Chapter 4 Computational Intelligence
1. Knowledge representation based on neural network (completed)
2. Genetic operator of genetic algorithm
Selection operators : roulette wheel selection and tournament selection ;
Crossover operator : one-point crossover and two-point crossover;
mutation operator .
3. The fitness of the roulette selection, the proportion of the fitness (completed)
The objective function f(x) may be a specific function given by the title. Observe whether the title says that the fitness is the same as the objective function value, or it is a mapping relationship. The proportion of fitness is the fitness in the total fitness proportion inside.
Chapter 5 Expert System
1. What is an expert system?
An expert system is a computer program system that simulates human experts to solve domain problems.
An expert system is a program system with a large amount of specialized knowledge and experience. It uses artificial intelligence technology and computer technology to reason and judge based on the knowledge and experience provided by one or more experts in a certain field, simulating the decision-making process of human experts. In order to solve complex problems that require human experts to deal with.
2. The main components of the expert system
- knowledge base
- Comprehensive database (global database)
- reasoning machine
- interpreter
- interface
3. Structure of rule-based expert system (completed)
4. Basic structure of neural network expert system (completed)
5. How does a rule-based expert system work? What is its structure?
A rule-based expert system is a computer program that uses a set of rules contained in the knowledge base to process specific problem information (facts) in the working memory and infer new information through the inference engine.
Its working model is shown in the figure ( completed ) :
Structure of a rule-based expert system:
- Knowledge base: build a human long-term memory model with a set of rules;
- Inference engine: build a human short-term memory model;
- Working memory: build a human reasoning model and infer new information;
- User interface (interface): the user observes the system through the interface and talks (interacts) with the system;
- Developer (person) interface: knowledge engineers develop the expert system through this interface;
- Interpreter: Provides an explanation for the reasoning of the system;
- External programs: such as databases, expansion disks and algorithms, etc., which support the work of the expert system.
6. Use a rule-based reasoning system to justify the following reasoning :
Known: Dogs can bark and bite
Any animal always bites when it barks
hounds are dogs
Conclusion: Hounds Bite
Step 1: The user imports the problem "hound dog" into the expert system through the user interface
Step 2: Based on the existing rules in the existing knowledge base, we can know
①Hounds are dogs, ②Dogs can bite
Step 3: Import the known rules ①② into the inference engine,
After reasoning, it can be seen that hounds are biting
Chapter 6 - Machine Learning
1. Definition of Machine Learning
Machine learning is a discipline that studies theories and methods for machines to simulate human learning activities, acquire knowledge and skills, and improve system performance.
2. Decision tree learning - training example (completed)
3. Convolutional Neural Network ( CNN )
(1) The convolutional neural network is essentially a mapping from input to output.
(2) CNN is a special deep neural network model, especially reflected in:
One is that the connection between its neurons is not fully connected
The second is that the connection between neurons in the same layer adopts the way of weight sharing .
(3) The basic structure of CNN includes two layers, namely feature extraction layer and feature mapping layer .
(4) Schematic diagram of convolutional neural network (completed)
4. What are the characteristics of reinforcement learning? What is the learning mode of the learning automata?
Features:
① Reinforcement learning is a trial-and-error learning (Trail-and-error). Since there is no direct guidance information, the agent must continuously interact with the environment to obtain the best strategy through trial and error.
② Delayed rewards, there is little guidance information for enhanced learning, and it is often given after the event (the last state), which leads to a problem, that is, how to distribute the rewards to the previous ones after obtaining positive or negative rewards state.
Learning Mode: (Complete)
Chapter 7 Intelligent Planning
1. Task planning (not completed)
2. The figure shows the world model where the robot works, and the robot is required to move the box from the initial position in room R2 to the target position in room R1. Try to establish the robot planning expert system, and give the planning results. (undone)
Prerequisites: state (robot, R1, box, R2) indicates that the robot is in room R1 and the box is in R2
Target state: state (robot, R1, box, R1)
Operation method (executable action)
move(1,2) means the robot moves from position 1 to position 2
hold (box) means pick up the box
drop (box) means put down the box
Step 1: The robot moves from R1 to R2 and picks up the box
Prerequisite: state(robot,R1,box,R2)
Operation method: move(R1, R2)
hold（box）
Step 2: The robot moves from R2 to R1 and puts down the box
Prerequisites: state(robot,R2,box,R2)
Operation method: move(R2, R1)
drop（box）
Reach the target state state (robot, R1, box, R1)
Chapter 8 - Natural Language Understanding
1. Research fields and directions of natural language understanding
Text recognition (Optical character recognition, OCR)
Speech recognition
Machine translation
Automatic summarization (Automatic summarization or automatic abstracting)
Syntax parsing
Text classification (Text categorization/document classification)
Information retrieval
information extraction
Information filtering
Natural language generation
Chinese word segmentation
Speech synthesis
Question answering system
2. Five levels of natural language understanding process
①Speech analysis
② Pragmatic analysis
③Sentence analysis
④ Semantic analysis
⑤ Lexical analysis