Bachelor - Artificial Intelligence Review Questions

Fill in the blank

1. First proposed the "artificial intelligence" in (1956) years.

2. The following is not part of the research methods of artificial intelligence (operation simulation) .

3. (Semantic network France) is a diagrammatic representation of knowledge of the structure, which consists of nodes and arcs or lines composed of chains, the structure can entity, and the causal relationship between physical properties and conciseness explicitly expressed.

4. In view of the search process, the rearrangement OPEN table means, whether in accordance with a (tentative value) re unexpanded node rearrangement.

5. The smaller the value of cost function, means that the node is located on the optimum route "hope" (larger) .

6. (Exploration heuristics) used to order nodes with extended, i.e., selection of the most promising node to be expanded so that the search efficiency is greatly improved.

7. In reasoning production system, from the bottom up is the reasoning ( forward chaining) .

8. Artificial intelligence, computational intelligence and biological intelligence from the low to high in the order of (computational intelligence, artificial intelligence, bio-intelligent)

9. The following does not belong to the main research areas of computational intelligence is (Network Computing)

10. The following does not belong to the main learning algorithm artificial neural network is a (non-enhanced learning)

11. A fuzzy inference is based on fuzzy logic based reasoning method uncertainty, is developed on the basis of two-valued logic syllogism, it (blur determination) as a precondition to (fuzzy language) rule, derive a Analyzing the approximate Fuzzy conclusions.

12, the definition of artificial life in 1987 as proposed by Rand:

Artificial life research is able to demonstrate the features of the behavior of natural systems artificial life system. Comprehensive study on the life of similar behavior by a computer or other machines.

13, the basic structure of the machine learning system is divided into four components, namely the environment, learning, knowledge base, perform, wherein (environment) learning portion to provide certain information to the system.

1. (Learning by analogy) is a useful and very effective method of reasoning, it is clear, concise description of the similarity between objects is an important method of human cognition of the world.

2. According to the definition of Fei Yade, databases (KDD) is identified from a large database of valid, novel, potentially useful, and can be understood by advanced process models.

3. The following does not belong to the method of knowledge discovery is the (smart calculation)

4. The key design expert system is (knowledge representation)

5. (Frame) is a structured representation, generally designated by the grooves of various aspects of things, each slot has a plurality of side surfaces, each side surface in turn has several values.

6. "If X does not know, then the conclusion Y", this definition belongs to (default reasoning)

7. The traditional school of logic and reasoning called classical logic and classical reasoning, logic and the new school and its reasoning known as non-classical and non-classical logical reasoning. Both compared on reasoning, using classical logic ( deductive logic ) reasoning, rather than using classical logic ( ILP ) reasoning.

8. There are three uncertain uncertainty reasoning, the following does not belong to these three uncertainty is (theoretical uncertainty)

9. (Genetic algorithms) is to mimic the biological mechanism of natural selection and genetics, one search algorithm by artificially constructed, it is a mathematical process of biological evolution simulation, evolution of calculating the most important form.

10. The new system consists of experts (Distributed Expert System, Cooperative Expert System)

11. The following features are not part of the expert system (timeliness)

12. There are four common deep learning model, respectively autoencoder, restricted Boltzmann machine, the depth of belief networks and convolutional neural network. Wherein (convolutional neural network) is essentially a mapping input to output, the basic structure of the model includes two layers, i.e., the feature extraction layers and features mapping layer.

13. Search into blind search and heuristic search 

14. The purpose is to allow the machine to artificial intelligence _ simulation, extension and expansion of human intelligence , in order to achieve some mental work mechanization.

15. In 1956 summer, a number of young scientists in the United States Dartmounth held a summer seminar in college, at that meeting, the first time the artificial intelligence of the term.

16. Three main ways for AI research: symbolism , behaviorism , connecting doctrine .

17. Blind search method has a depth-first , breadth-first   and  backtracking strategy and so on .

18. Genetic manipulation simple genetic algorithm mainly selection , crossover and mutation .

19. Artificial neural networks have the basic properties: nonlinear   , nonlocal   , unsteady   and non-convexity .

20. The main application areas of artificial intelligence are: natural language understanding    ,   intelligent retrieval database    ,  expert advisory system , theorem proving  , games , robotics   , automatic programming , a combination of scheduling problems   , perception problems and so on.

21. Uncertainty type divided by nature: random, fuzzy, incomplete, inconsistent .

22. In the Maze, board games, eight games, digital games, commonly used artificial intelligence to the core technology called graph search time technology, the solution of these problems, and often put in the position of the labyrinth, the layout of chess, the eight digital Fig situation arranged to form, this graph is called the state space of FIG  .

23. In the famous medical expert system MYCIN stipulates that if the credibility of evidence A CF (A) = 0, it means nothing to do   , CF (A) = - 1 , it means  false    , CF (A) = 1, it means the truth .  

24. There are three uncertainties _ Uncertainty Reasoning uncertainty about the evidence , on the uncertainty of knowledge and _ uncertainty about the conclusions .  

25. The performance characteristics of human intelligence in four areas can sense the objective world of information , can be thought of by the knowledge acquired for reprocessing , through the accumulation of knowledge and improve their ability to learn and adapt to environmental change and can respond to external stimuli transmission of information

26. Artificial neural networks can be divided into recurrent network and a multi-layer network two basic structures.

27. Particle swarm optimization algorithm is an algorithm-based search group, which is based on the simulated birds society.

Complex and short answer questions

1. In the left bank of the river there are three monks, three savage and a boat, monks want to use the boat to all the people across the river, but subject to the following conditions:

   (1) monks and Savage will be rowing, but the ship can only transport a maximum of two people;

   (2) In any savage shore shall not exceed the number of monks, or monks 

        Savage will be eaten.

   Savage assumed would obey any kind of arrangement to cross the river and try to plan out a way to ensure the safety monk river program.

answer:

1 ) state the problem

It can be described by a triplet:

           S＝(m, c, b)

       m : number of left bank monk

       c : The number on the left bank of the wild

       b : the number of vessels on the left bank of

     The right bank of the state does not have to mark, because:

       Monk right bank number m '= 3-m

       The number of the right bank of wild c '= 3-c

       The number of vessels on the right bank b '= 1-b

status	m, c, b	status	m, c, b	status	m, c, b	status	m, c, b
S0	3  3  1	S8	1  3  1	S16	3  3  0	S24	1  3  0
S1	3  2  1	S9	1  2  1	S17	3  2  0	S25	1  2  0
S2	3  1  1	S10	1  1  1	S18	3  1  0	S26	1  1  0
S3	3  0  1	S11	1  0  1	S19	3  0  0	S27	1  0  0
S4	2  3  1	S12	0  3  1	S20	2  3  0	S28	0  3  0
S5	2  2  1	S13	0  2  1	S21	2  2  0	S29	0  2  0
S6	2  1  1	S14	0  1  1	S22	2  1  0	S30	0  1  0
S7	2  0  1	S15	0  0  1	S23	2  0  0	S31	0  0  0

2 ) operation set F. = {P01, P10, P11, P02, P20, Q01, QlO, Q11, Q02, Q20}

3 ) the state space

After the description given and the operation state, the state space of the problem are:

{{S0}，{ P 01，P 10，P 11，P 02，P 20，Q01，Q 10，Q 11，Q 02，Q 20}，{S31}}。

4）状态空间图：

 

 

 

 

 

 

 

 

 

 

2.

3.

4.

5.

6.

7.

8.人工智能中的知识表示方法有几种？

1）状态空间表示

2）问题归约表示

3）谓词逻辑表示

4）语义网络表示

5）框架表示

6）过程表示

9.

10.

11.人工智能的研究目标

12.人工智能发展时期？

1孕育时期（1956年前）

2形成时期（1956——1970年）

3暗淡时期（1966——1974年）

4知识应用时期（1970——1988年）

5集成发展时期（1986年至今）

13.

14.人工智能研究的基本内容

  认知建模

  知识表示

 知识推理

  知识应用

  机器感知

机器思维

  机器学习

  机器行为

  智能系统构建

15.遗传算法的求解原理和步骤？

原理: 遗传算法是模拟达尔文生物进化论的自然选择和遗传学机理的生物进化过程的计算模型，是一种通过模拟自然进化过程搜索最优解的方法。遗传算法是从代表问题可能潜在的解集的一个种群开始的，而一个种群则由经过基因编码的一定数目的个体组成。每个个体实际上是染色体带有特征的实体。染色体作为遗传物质的主要载体，即多个基因的集合，其内部表现（即基因型）是某种基因组合，它决定了个体的形状的外部表现，如黑头发的特征是由染色体中控制这一特征的某种基因组合决定的。因此，在一开始需要实现从表现型到基因型的映射即编码工作。由于仿照基因编码的工作很复杂，我们往往进行简化，如二进制编码，初代种群产生之后，按照适者生存和优胜劣汰的原理，逐代演化产生出越来越好的近似解，在每一代，根据问题域中个体的适应度大小选择个体，并借助于自然遗传学的遗传算子进行组合交叉和变异，产生出代表新的解集的种群。这个过程将导致种群像自然进化一样的后生代种群比前代更加适应于环境，末代种群中的最优个体经过解码，可以作为问题近似最优解。

步骤:

16.遗传算法的执行过程？

 

17.消解原理？

18.消解推理规则？

19.消解反演

20.反演求解过程？

从反演树求取对某个问题的答案，其过程如下:
(1)把由目标公式的否定产生的每个子句添加到目标公式否定之否定的子句中去，

21.神经网络

22.对某种产品的质量进行抽查评估。现随机选出5个产品x1,x2,x3,x4,x5进行检验,它们质量情况分别为:

x1＝80,x2＝72,x3＝65,x4＝98,x5＝53

这就确定了一个模糊集合Q,表示该组产品的“质量水平”这个模糊概念的隶属程度。试写出该模糊集。

答案：

该模糊集为：Q={(X1,0.8),(X2,0.72),(X3,0.65),(X4,0.98),(X5,0.53)}

 

 

23.

24.