New Semester
Started
Get
50% OFF
Study Help!
--h --m --s
Claim Now
Question Answers
Textbooks
Find textbooks, questions and answers
Oops, something went wrong!
Change your search query and then try again
S
Books
FREE
Study Help
Expert Questions
Accounting
General Management
Mathematics
Finance
Organizational Behaviour
Law
Physics
Operating System
Management Leadership
Sociology
Programming
Marketing
Database
Computer Network
Economics
Textbooks Solutions
Accounting
Managerial Accounting
Management Leadership
Cost Accounting
Statistics
Business Law
Corporate Finance
Finance
Economics
Auditing
Tutors
Online Tutors
Find a Tutor
Hire a Tutor
Become a Tutor
AI Tutor
AI Study Planner
NEW
Sell Books
Search
Search
Sign In
Register
study help
business
management and artificial intelligence
Artificial Intelligence In The 21st Century 2nd Edition Stephen Lucci, Danny Kopec - Solutions
1. What is the relationship between annealing and SA?
10. Write a program to solve the TSP when n = 10 cities, using a Discrete Hopfield network(consult Exercise #15). As in the previous program, run your program 10 times and discuss your results.
9. Write a program to solve the 4–Queens problem using a Discrete Hopfield Network (consult Exercise #14). Run your program 10 times choosing different units for update on each run.Discuss your results. (Recall that the Hopfield network finds a local energy minimum, and therefore you should
8. Use your program in Programming Exercise 5 to classify irises into three classes—Setosa, Versicolor, and Virginia. [See DVD, Appendix D.2.1]
7. Use your program in Programming Exercise 5 to predict the next week’s gold price based upon the following data. Use the last 25% of the data for validation. [See DVD, Appendix D.2.1]
6. Use your program in Programming Exercise 5 to approximate the function that provides the weight in mg of Wild Australian Rabbits as a function of age (in days). Withhold every third data item in this table for validation purposes. [See DVD, Appendix D.2.1]
5. Write a program that can apply the backpropagation algorithm to any two-layer feed forward network.
4. Write a program to implement the backpropagation algorithm to train the two-layer network in Figure 11.38 for the XOR function.
3. Use the Delta Rule to complete the training for the two-input OR function in Example 11.2.
2. There are 24 or 16 Boolean functions of 2 variables. Use the Perceptron Learning Rule to determine how many of these are linearly separable.
1 and ten lie in Class 2 and these classes are linearly separable. Use the Perceptron Learning Rule on this data with a learning rate α = (a) 0.01 (b) 0.1 (c) 0.25 (d) 0.50 (e) 1.0 (f) 5.0.Comment on the performance of the learning algorithm in each case.
1. Generate 20 triples of random numbers (60 in total) where each number ∈ [0, 1]. Each triple corresponds to a point in the unit cube. Generate these numbers so that ten triples lie in Class
16. Consider the Hopfield network drawn in Figure 11.67. Calculate the energy for each state and draw the state transition diagram for this network. Identify the stable states (if any).
15. The TSP was also discussed extensively in both Chapters 2 and 3. Specify the architecture of a Discrete Hopfield Network to solve a small instance of this problem, say n = 4 cities.Hint—use an n × n Boolean matrix to represent your tour. A “1” should be placed in column j of row i if
14. The n-Queens Problem was discussed extensively in Chapter 2. Specify the architecture of a Discrete Hopfield Network to solve this problem when n = 4.
13. The n-Rooks Problem is to place n rooks on an n × n chessboard so that these pieces are nonattacking. A rook can attack any piece on the same row or column. A solution to the 4–rooks problem is shown in Figure 11.66.Specify the architecture of a Discrete Hopfield Network to solve this
12.a. Propose a design for a BPN to help make health insurance premium decisions for an insurance company. Prospective clients for health insurance are to be classified as lowrisk or high-risk candidates, where high-risk clients are to be charged a higher premium or may even be refused insurance.
11.a. Solve Exercise 6 using the Delta Rule. Train the neuron for one epoch.b. What is the stopping criterion for this learning rule?
10. Which of the following sets of points are linearly separable?a. Class 1: {(0.5, 0.5, 0.5), (1.5, 1.5, 1.5)}Class 2: {(2.5, 2.5, 2.5), (2.5, 2.5, 2.5)}b. Class 1: {(1, 1, 0), (2, 3, 1), (3, 2, 1.5)}Class 2: {(1, 1, 2), (2, 3, 2.5), (3, 2, 3.5)}c. Class 1: {(0, 0, 18), (2, 1, 10), (7, 5, 4)}Class
9. Suppose that you have a single TLU with n inputs; in other words, unaugmented input vector xi , will have n components, and you are performing Perceptron Learning. How many epochs ust you wait before you can be certain that the algorithm will not halt (that is, that the patterns are not linearly
8. A TLU is being trained using the Perceptron Learning Rule. Input vector and weight vector appear as in Figure 11.65.The target t for the current pattern = 1, however, the actual output of the unit y equals 0.Which way should w be rotated with respect to x ? Explain your answer.
7. Use the Perceptron Learning Rule to learn the majority function on three inputs where the second input x2 is held fixed at 1. Maj(x1,x2,x3) = 1 whenever two or three of x1, x2, and x3 are equal to 1, then Maj(x1,x2,x3) = 1. All inputs are binary and the initial weight values are( 1 2 3 ), , , 3
6.a. Use the Perceptron Learning Rule to train a neuron to learn the two-input function depicted in Figure 11.64. Use an augmented input vector. Let the initial weight values be w1 = 0.1, w2 = 0.4 and θ = 0.3. Use a learning rate α = 0.5.b. Give the equation of the discriminant and draw this line
5. Prove that the two-input XNOR function cannot be implemented with a single perceptron. Your proof should use a system of inequalities.
4. It is a well-known physiological phenomenon that if a cold stimulus is applied to a person’s skin for a very short period of time, the person will perceive heat. However, if the same stimulus is applied for a longer period, the person will perceive cold. The use of discrete time steps enables
3. What function F is computed by the McCulloch-Pitts network in Figure 11.62?
2. Design a McCulloch-Pitts network for the three-input minority function where Min(x1,x2,x3) equals 1 whenever only one or none of the inputs equal 1, in other words, Min(x1, x2, x3) = x1′x2′x3′ + x1′x2′x3′ + x1′x2x3′ + x1x2′x3′.
1. Draw a McCulloch-Pitts network to implement the sum function S for a full adder where S(ABCi) = A′B′Ci + A′BCi′ + AB′Ci′ + ABCi.
13. Contrast supervised and unsupervised learning.
12. What are the differences between biological and artificial neurons (neural networks) in terms of both structure and functionality?
11. The human brain consists of between 10 billion (1010) and 100 billion (1011) neurons. Once we understand the workings of the human brain 1 and we construct full-scale software and/or hardware simulations, what do you predict will occur?Kurzweil predicts (Kurzweil 1999) that this will occur by
10. What is the difference between offline and batch training?
9. Why will both the Delta Rule and backpropagation continue to have some error, whereas the Perceptron Learning Rule halts when there is no error present?
8. The backpropagation algorithm is often referred to as the generalized Delta Rule. Why do you think this is so?
7. What information does the dot product of x with w provide in an ANN? How is this information used in the following:a. The Perceptron Learning Rule?b. The Delta Rule?c. Backpropagation?
6. The learning rate is a constant between 0 and 1, in other words, 0 < α ≤ 1. Since a larger learning rate results in faster learning, why not use large values for α?
5. A single-layer neural network cannot implement a function that is not linearly separable. Is this a serious drawback? Explain.
4. Why might stress in humans be considered a nonlinear phenomenon?
3. Nonlinear systems do not obey the proportionality relation between input and output changes as do linear systems. Consider an artificial neuron with a threshold θ = 0.50. Argue that this neuron is a nonlinear system.
2. In a linear system, the output is proportionally related to the input, in other words, small changes in the input produce correspondingly small changes in output and similarly for large input changes. Describe two systems in nature that are examples of linear systems.
1. In Section 1 we portray an ANN as a black box. What limitations does this opaqueness impose on their utility?Next Page
6. If your decision tree does not converge, what is needed here?i. More input data?ii. Attributes that do a better job of separating hypotheses?
5. Refer first to exercise 3. Design a decision tree to distinguish between bronchitis, pneumonia, and TB.
4. Table 10.2 contains data about two medical conditions:A person has a cold vs. a person has the flu. Use ID3 to build a decision tree to determine which affliction a person suffers from based upon their symptomology.
3. Test your trees from programming exercises 1 and 2 with:Spaghetti carbonara where sauce = white, contains meat = true, and contains seafood is false.What result did you obtain? Is it what you expected? Explain.
2. Add some noisy data to the tree construction program above.Comment on what happens. For example:Sauce = red, contains meat = true, and contains seafood = false, but likes = no.Penne with Bolognese sauce would yield these attribute values.Go on-line to find several other “noisy” examples.
1. Use ID3 to confirm the final form of the decision tree for pasta preferences in Section 10.7.
3. Entropy is also defined when classification is into three or more classes.The entropy of a set S relative to n distinct classes is defined as:Entropy (S) = 2 1 log n i i i p p = Σ−where pi is the proportion of the Set S in class i, with i = 1, …, n. Notice that logarithms are still
2. Calculate the entropy for each of the following sets:a. [6(+), 11(-)]b. [1(+), 9(-)]c. [2(+), 12(-)]
1. Design a decision tree for the following Boolean functions:a. a ˅ (b ˄ ~ c)b. majority (x,y,z)
13. Are decision trees a lazy or eager learner? Explain your answer.
12. Suggest a possible method to handle attributes with continuous values.
11. Why does choosing that attribute with the largest information gain favor construction of shorter decision trees?
10. When calculating the entropy of a set, why are logarithms calculated with a base equal to 2?
9. Look on the web and find several other areas where decision trees are used.
8. Give an example of where you have used Occam’s Razor in your everyday experience.
7.a. What is Occam’s Razor?b. Does it claim that the shortest hypothesis is always the best?
6. When performing curve fitting, why is a function that passes through all the points in a training set not necessarily the best hypothesis?
5. Describe inductive learning.
4. Why is feedback important to an agent?
3. Describe three different forms of feedback in ML systems.
2. List several ML paradigms.
1. What is machine learning and why is it such an important subfield of AI?
15. Research expert systems built in the past 5 years. What are their features?How are they different from earlier expert systems described in this chapter?
14. How would expert systems need to perform to pass the Turing Test?
13. One of the criticisms of expert systems has been that they are conducive to the creation of microworlds (e.g. by Professor Hubert Dreyfus, see Chapter 6, Section 6.8, p.185). Explain why you do or do not agree.
12. Who owns expert systems? Expert systems have long been considered a major success story from the field of artificial intelligence; however, they have also become somewhat standard and common. Should expert systems be considered a computer science technology or do they strictly belong to AI?
11. Explain why MYCIN was such an important program to all future expert systems and shells.
10. Explain the difference between procedural knowledge, declarative knowledge, and metaknowledge.
9. Explain why expert systems are different from conventional programs?
8. Why is it important for an expert system to be cost effective?
7. Why is a procedure such as the Rete Algorithm important to expert systems development?
6. Do you believe expert systems can outperform human experts? If not, explain why not. If so, give some examples and describe what these can do that a human expert cannot do.
5. What is a Demon Rule? Develop prototype demon rules for your expert system.
4. Did you use/need a domain specialist for your problem domain? If not, consider how a domain specialist might help you. If so, consider the knowledge engineering process that took place between you and the domain specialist.
3. Evaluate your system. How good is its performance? How could it be improved? Can it be used as a practical tool?
2. Try to build an expert system using CLIPS in your domain of interest.
1. Consider a domain you might want to build an expert system for. What characteristics should the domain possess in order to be a good candidate?
21. Describe some of the hybrid intelligent techniques used to build these systems.
20. Name three expert systems built in the last decade and their application domains.
19. Describe several problems in building a CBR?
18. Name four typical aspects of constructing a CBR System?
17. What is the idea behind case-based reasoning (CBR)?
16. What is the Rete Algorithm?
15. What is a Demon Rule?
14. Why was MYCIN such an important program?
13. Describe the main purpose and main method of DENDRAL.
12. Explain why knowledge acquisition is the “bottleneck for AI.”
11. Describe the process of knowledge engineering.
10. Name 5 different expert systems in five different areas.
9. List 10 application areas for expert systems.
8. List 10 purposes expert systems have been built for.
7. Describe 10 characteristics of expert systems.
6. Describe 10 characteristics of human experts.
5. What was Dreyfus & Dreyfus’s main position on the limitations of AI?
4. Describe the Five Stages of Skill Acquisition.
3. Explain how humans can compete with programs in performance of activities in which programs perform millions of computations.
2. Explain how a domain specialist might have some 50,000 concepts in his skill domain.
1. Explain how expert systems fit into the advances in technology in the time during which they were developed.
8. A Dutch Book is some mix of wagers, which can be shown to result in a sure loss (according to the bettor’s beliefs). Consider this situation (The three-card problem): there are three cards, one is red on both sides (RR), the second is red on one side and white on the other (RW), and the third
Showing 3500 - 3600
of 4756
First
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
Last
Step by Step Answers
Get 50% OFF
Claim Now
