Thinking As Computation A First Course 1st Edition Hector J Levesque - Solutions

Discover comprehensive solutions for "Thinking As Computation: A First Course" by Hector J Levesque, designed to enhance your learning experience. Our online platform provides a detailed answers key, offering step-by-step solutions to help you grasp complex concepts. Access the solution manual and download the solutions PDF for free, ensuring you have solved problems at your fingertips. Our extensive test bank includes chapter solutions, questions, and answers, making it an essential resource for students and instructors alike. With the instructor manual and textbook solutions readily available, you can confidently tackle any challenge in your studies.

2. During back-propagation network training, is it generally desirable to train the network to have as small a sum-squared error as possible on the training patterns?Why?
1. In back-propagation networks, why should we use PEs with biases in the hidden and output layers? Why choose nonlinear functions as activation functions?
9. Assume we want to scale inputs to [-1, 1] for z-axis normalization. One of the input parameters varies over all the patterns from -4.2 to + 10.0. How would you scale this input? Why?
8. Derive a back-propagation (BP) adaptation algorithm for a four-layer BP neural network. Assume the activation function of the hidden PEs is a sigmoid function as expressed by equation 5.23. The activation function of the output PEs is a linear function as expressed by equation 5.8 with a = 1.
7. Prove the convergence of the binary Hopfield network.
6. Review White (1989). Summarize the reasoning behind the proof that one hidden layer is sufficient to approximate virtually any nonlinear function.
5. What are the differences among supervised adaptation, unsupervised adaptation, and reinforcement adaptation?
4. If one hidden layer of sigmoidal PEs can approximate any nonlinear function, why might we decide to use more than one?
3. The sigmoid activation function is 1/(1 + e-input). Derive the first derivative of this function.
2. Show that a hidden layer doesn't change (improve) network performance if all PEs (hidden and output) have linear activation functions.
1. In a single-layer neural network with n processing elements (PEs), how many unique weights are possible if the only restriction is that no self-feedback connections are allowed? How many are possible if it is also specified that weights are symmetric, that is, wji = wij?
10. Run the PSO implementation using two swarms to optimize the G1 function. Try two different population sizes for each of the two swarms (four combinations total). Turn in and discuss your results.
9. Run the PSO implementation as a single swarm, optimizing the F6 function. Note that the run file example in Listing 4.36 is for two swarms, so make sure you have an appropriate run file for a single swarm. Try two population sizes. Turn in and discuss your results.
8. Run the GA implementation, optimizing the Griewank function. Try two different crossover types and two different selection types (four combinations of parameters). For each combination of parameters, how many total generations are required to achieve a fitness of -1.3 or better? Turn in and
7. Implement the benchmark function you identified in exercise 5 into the PSO source code.
6. Add the capability for specifying a termination criterion (acceptable error, for example) to the GA source code.
5. Five benchmark functions are identified in Table 4.2. Identify an additional benchmark function appropriate for evolutionary algorithms. Justify your choice. Modify the source code for the GA to implement this benchmark function so that it becomes an additional choice for the user.
4. Draw a state diagram for the synchronous version of PSO and compare it with the asynchronous version.
3. Draw a flowchart for the implementation of PSO.
2. Draw a state machine diagram for the GA implementation.
1. In the implementation of mutation operator, a "mutation according to bit position"flag is used to tell whether or not a mutation by bit position is implemented.Define an enumeration data type to replace the flag and make corresponding changes in the implementation of the mutation operator.
11. How is a particle swarm optimizer similar to a genetic algorithm? How is it different? How does it resemble an evolution strategies implementation?
10. Sketch out a genetic programming representation of the best possible approxi- mate solution to v = rh, (v is the volume of a right cylinder, r is its radius, and h is its height) given that the maximum depth of the program is five layers and you may only use the constant values 0, 1, and 10. If
9. Assume you are going to use genetic programming to evolve a program to classify the Iris dataset (pp. 197-198). Specify a function set and a terminal set that are appropriate to solve the problem.
8. What is the main difference between evolutionary programming and evolution strategies?
7. Assume each population member in a GA consists of 8 binary coded bits (as in the GA example in the chapter), representing the integers 0 to 255. Briefly describe or sketch the portion of the problem space covered by the following schemata: 0*******, *******1, 10******, ****** 10, ***11****
6. Assume that the average fitness of strings containing a particular schema S is 20 percent less than the average fitness of all schemata, and the schema appears in 50 percent of the initial population. Assume that the probability of disrup- tion of this schema by crossover or mutation is
5. After running a genetic algorithm for a fairly long time, the fitness values tend to cluster at the high end of the scale. For example, on a scale of 0 to 1, they might cluster from 0.90 to 0.98. What is the main problem with this? How can it be alleviated?
4. Assume standard binary encoding of parameters is used for a genetic algorithm implementation. Briefly discuss how the effects of uniform crossover and two-point crossover change as the number of bits representing a parameter is increased.
3. According to the schema theorem, what happens to highly fit schemata in successive generations? What are the effects of crossover and mutation according to the theorem? Why use crossover and mutation?
2. How many schemata are possible for a 6-bit binary string?
1. Convert the following binary coded strings to Gray coding: 10101010, 10011100, 01100110.
6. Give a real-world example of each type of adaptation: supervised, reinforcement, and unsupervised.
5. Randomness is sometimes listed as an attribute of intelligent systems. Why?
4. Find another source of information on cellular automata. Discuss the relationship between cellular automata and self-organization.
3. Find an article or a chapter in another book on emergent computing. Compare the concept of emergent computing as presented there with the concept of self-organization presented in this chapter.
2. Read other discussions of computational intelligence, including Bezdek (1998). Develop your own one-paragraph definition of computational intelligence.
1. What other elements might be appropriate for inclusion in the world model of Figure 2.5?
What is the definition of artificial intelligence? List some differences between computational intelligence and artificial intelligence.
What is the difference between fuzziness and probability? Provide an example to illustrate the difference.
Survey recent technical publications and the Internet for these additional areas to which one of the component technologies of CI has been successfully applied: face recognition, health screening, creating art.a. What motivated the use of the technology in these applications?b. What technical
How do you think adaptation and self-organization are interrelated?
State a myth relative to neural networks, fuzzy systems, or evolutionary computation, in addition to those discussed in this chapter. Why is it a myth?
What are some alternative terms for processing element? Discuss the choices, listing advantages and disadvantages for each.
Can Computers Really Think?
9. The game player presented in the file minimax.pl can find a move to guarantee a given numeric value, if one exists. However, it cannot find the best move to make, since it does not compare the values produced by all possible moves. Write a game player that does this. Hint: Use the built-in
8. Look on the Internet to discover the strange connection between winning strategies for Wythoff’s Nim and the golden mean, that is, (−1+√5)/2, the remarkable number studied since the time of the ancient Greeks.
7. Use exercise 6d to describe a winning strategy for Wythoff’s Nim. Given an arbitrary state of the game [p1, p2], what is the best move to make?
6. Consider the following game, called Wythoff’s Nim, a variant of Race to 21:When the game begins, there is a pile of green chips and a pile of red chips on the table. Each player takes turns removing at least one chip from the table, and the last player to take a chip wins. In a move, a player
5. Consider the numeric game tree to depth 4 in figure 10.21.a. Determine the numeric values for all the nodes in the tree.b. Indicate on the tree where alpha and beta cutoffs would occur, assuming that the tree is explored in the usual way, from left to right. Of the entire tree, how many leaf
4. One of the reasons the Prolog general game player presented in the file gameplayer.pl may be hard to follow is because of the possibility of a tie. Write a simpler version of label and win_move (and any other predicate you need) for games where there is always a winner. Show that your program
3. This question concerns the intuitive understanding of the labels on a game tree.Suppose players P and Q are playing a game, and each state of the game is labeled with P or Q (as the eventual winner) or neither for a tie.a. If a state is labeled with P as the winner, which of the following
2. Consider the game tree depicted in figure 10.20. What is the move that Max should make from the root node? Answer by labeling every state of the game tree with W, L, or T (for a tie).
1. Consider a game of tic-tac-toe where X begins by choosing position 2.a. Show that the general game player in gameplayer.pl finds no winning move for O from here.b. The usual move for player O here would be a center or a corner. Show that the general game player finds an equally good move for O.
9. Consider the following, called the jealous husband problem:Three married couples find themselves on one bank of a river wanting to get to the other side using a boat that can only hold two people at a time.However, each husband insists on being with his wife whenever there are other men
8. As noted in section 9.3.2, the reachable predicate defined in figure 9.3 allows for loops, where a sequence of moves passes through a state and later returns to exactly the same state. For example, in the monkey and bananas problem where the monkey starts at location 2, a sequence of go moves
7. This exercise returns to a blocks world similar to the one shown in figure 8.16.Imagine that there is a group of blocks (cubes, for simplicity) arranged in some configuration on a table in front of a robot, and you would like the robot to rearrange them without having to tell it how to do so.
6. The bplan predicate presented in figure 9.8 can run forever, looking for longer and longer plans, if it happens that there is no way to achieve the goal. Modify the predicate so that it takes an extra argument n and will eventually stop with failure if there is no plan of length n or less.
5. Consider the reachable predicate defined in figure 9.3. Suppose the second clause had been this:reachable(S1,[M|L],S3) :-reachable(S2,L,S3), legal_move(S1,M,S2).Are there problems where this might work better than the previous version? Are there problems where it might do worse?
4. Consider the following well-known problem:In his yard, a farmer has a fox, a hen, and a bushel of grain. The farmer must transfer all these things to a stall at the market, using only a cart which can carry at most one of them (plus himself) at a time. The problem is that if the farmer leaves
3. Consider a more ambitious version of the monkey and bananas problem. In this version, the monkey cannot reach the bananas even if he is on the box. However, there is a big stick located in a corner of the room that can be used to knock the bananas down when the monkey has the stick and is on the
2. Suppose the plan predicate from figure 9.3 had been defined by plan(L) :- initial_state(I), reachable(I,L,G), goal_state(G).Are there planning problems where this might work better than the previous version? Are there problems where it might do worse?
1. The version of the monkey and bananas problem in section 9.2.4 did not consider what happened after the monkey grabbed the bananas. Imagine that there is a very clever monkey who wants to trick the scientists studying him (while they are out having coffee) by obtaining the bananas as before, but
10. Handle the article the properly. The idea here is that a noun phrase like “the red pyramid” should only succeed in naming a block if there is a unique block of the appropriate kind.
9. Show that your grammar for imperative sentences is ambiguous by drawing two parse trees for the third example sentence in exercise 7. Also show that your parser can sometimes disambiguate a sentence using facts about the current scene.
8. Test your parser on a variety of imperative sentences like those in exercise 7, including some that are syntactically well formed but impossible to execute.
7. Extend your parser/interpreter from exercise 5 to handle simple imperative sentences of the form “Put NP on NP” where NP is a noun phrase. For example, the following are syntactically well-formed imperative sentences:Put a green pyramid on the small red cube.Put an orange wedge on the block
6. Write clauses defining the put_on predicate, and test it on a variety of move operations. After each operation, include a sketch of what the new scene looks like, and have your program answer some queries involving that new scene.
5. Copy the Prolog parser/interpreter for noun phrases presented in the chapter(or write your own). Then test the np predicate on a variety of noun phrases to show that it is capable of identifying the blocks being referred to in your scene.To model moving a block within your blocks world, suppose
4. Build a Prolog lexicon of articles, adjectives, common nouns, and prepositions, including all the words in the six example noun phrases. The word any should be treated as an article.
3. Show that your world model works properly by getting Prolog to answer some queries about it. (These queries should not use English noun phrases.)Once your world model is working satisfactorily, you are ready to consider English noun phrases and the blocks they refer to in your scene. Here are
2. Write clauses defining the predicates beside(x, y), which holds when block x and y are both on the table in two adjacent areas, and above(x, y), which holds when block x is somewhere above block y.
1. Build a world model that expresses basic facts about colors, sizes, shapes, and locations of some blocks in a scene. Include within your documentation a drawing of the blocks-world scene that corresponds to that world model.
6. Write Prolog clauses that define each of the following predicates:a. exactly_3(x) is true if x is a list with exactly three elements.b. at_least_3(x) is true if x is a list with at least three elements.c. at_most_3(x) is true if x is list with at most three elements.Hint: Use four clauses.d.
5. The list version of the blocks world gave short definitions of the above and left predicates. Define the on predicate using the just_before predicate.
4. Use the append predicate to define a predicate just_before(x, y, z) that holds when x appears just before y as an element of list z. So, for example,?- just_before(3,4,[1,2,3,4,5,6]).Yes?- just_before(3,5,[1,2,3,4,5,6]).No Hint: This can be done with one clause. Look at how elem2 is defined.
3. Explain informally why [X,[1,X,3]] does not unify with [Y,Y]. It turns out that Prolog thinks these two lists do unify: it creates a cyclic term, which in this case is a list that contains itself as an element.
2. Consider the mystery predicates p and q defined by the following three clauses:p(X) :- q(X,[],X).q([],Y,Y).q([E|L],Y,Z) :- q(L,[E|Y],Z).Describe in English the lists x for which the predicate p(x) holds.
1. The text defined a predicate elem(x, y) that held when x was an element of the list y. Consider the following variant:elem3(X,[X]).elem3(X,[_|L]) :- elem3(X,L).Give an example of an x and a y such that elem(x, y) holds but where elem3(x, y)does not hold. Describe in English when the predicate
7. Consider a course timetable for a Fine Arts student at a mythical university.The student wants to take a course in art, music, film, and dance, and each of these has three hours of classes a week. To accommodate a variety of timing constraints, some courses offer two sections at different
6. The answers submitted by five students to a true/false quiz are as follows:Teresa: true, true, false, true, false Tim: false, true, true, true, false Tania: true, false, true, true, false Tom: false, true, true, false, true Tony: true, false, true, false, true Tania got more answers right than
5. Many newspapers feature a puzzle called KenKen in the recreational section.The KenKen puzzle is similar to Sudoku except that arithmetic relations are involved. Write a Prolog program that solves a 4 × 4 KenKen puzzle taken from a newspaper.Hint: You will need the uniq predicate used for the
4. Consider the following problem:Donna, Danny, David, and Doreen were seated at a table in a restaurant.The men sat across from each other, as did the women. They each ordered a different main course with a different beverage. In addition,– Doreen sat beside the person who ordered steak.– The
3. Use Prolog to solve the cryptarithmetic puzzle CROSS + ROADS = DANGER.Note: If you are not careful with the ordering of constraints in your program, the program could run for hours. With a proper ordering, it should only take a second or so. Explain briefly the order you have chosen.
2. Draw a map of central Europe consisting of the following seven countries: France, Switzerland, Italy, Belgium, Holland, Germany, and Austria. Color the countries on the map using just red, yellow, and orange; no two countries with a border between them can be the same color. Look at an existing
1. This chapter presented three different ways of handling equality in Prolog: =,=:=, and is. For each of the three, explain briefly what its purpose is, and present an example setting where it should be used but where the other two should not be used.
10. write_descendant_chain.Later chapters examine the special predicates write and nl, which can be used to produce output other than just the values of variables. For example, if there is a clause write_child(X,Y) :-write(X), write(’ is a child of ’), write(Y), nl.then the query
9. closest_common_ancestor.x is a closest common ancestor of two people y and z if x is an ancestor of both y and z and no child of x is an ancestor of y and z. (A closest common ancestor of two first cousins will be someone who is a grandparent of both.) These exercises will give you practice
8. cousin.x is a cousin of y if some parent of x and some parent of y are either siblings or (recursively) cousins. (The x and y will always end up being nth cousins for some n. For example, fourth cousins will have parents who are third cousins.) These exercises will give you practice writing and
7. descendant, ancestor.x is a descendant of y if x is either a child of y or (recursively) of someone who is a descendant of y. x is an ancestor of y if y is a descendant of x. These exercises will give you practice writing and running Prolog programs that reason in a simple way about family
6. first_cousin_twice_removed.Look on the Internet or elsewhere for the definition of a third cousin once removed or a second cousin twice removed and so on, and fill in the details here. These exercises will give you practice writing and running Prolog programs that reason in a simple way about
5. half_first_cousin and double_first_cousin.x is a half first cousin of y if some parent of x is a half sibling of a parent of y. x is a double first cousin of y if each parent of x is a sibling of a parent of y. (Note that a first cousin is not the same thing as a double half first cousin. The
4. first_cousin and second_cousin.x is a first cousin of y if some parent of x and some parent of y are siblings. (Note that first cousins will have a grandparent in common.) x and y are second cousins if some parent of x and some parent of y are first cousins. (Second cousins will share a
3. half_sibling and full_sibling.Half siblings are siblings who have exactly one parent in common. Full siblings are siblings who have two parents in common. These exercises will give you practice writing and running Prolog programs that reason in a simple way about family relationships, mostly
2. sibling, brother, and sister.x is a sibling of y if x and y are two different people who share a parent in common. A brother is a sibling who is male; a sister is a sibling who is female. These exercises will give you practice writing and running Prolog programs that reason in a simple way about
1. mother, grand_parent, and great_grand_mother. These exercises will give you practice writing and running Prolog programs that reason in a simple way about family relationships, mostly exotic flavors of cousins. Begin your Prolog program by writing a collection of clauses for the child, male, and
Make up a simple knowledge base of atomic sentences about movies (not necessarily real ones) using the following predicates:acted_in(person,movie) the person acted in the movie directed(person,movie) the person directed the movie released(movie,year) the movie came out that year You can use numbers
5. Suppose the (incorrect) atomic sentence yonge is left of bay.were added to the original knowledge base. For what values of X would the resulting knowledge base entail spadina is west of X? What happens now with the back-chaining procedure on the query (3f) bay is west of sherbourne?
4. Suppose the second conditional sentence in the knowledge base is replaced by If X is west of Y and Y is left of Z then X is west of Z.Would this change what is entailed? What happens now with the back-chaining procedure on the query (3d) yonge is west of yonge?
3. Trace the back-chaining procedure on the following queries:a. spadina is west of st-george;b. yonge is east of bay;c. christie is west of spadina;d. yonge is west of yonge;e. st-george is east of bathurst;f. bay is west of sherbourne.

Showing 3200 - 3300 of 4756

First
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
Last

Thinking As Computation A First Course 1st Edition Hector J Levesque - Solutions

Step by Step Answers