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

Result Not Found

Books FREE
Study Help
Tutors
AI Study Planner NEW
Sell Books

Search
Sign In
Register

study help
business
management and artificial intelligence

Artificial Intelligence Foundations Of Computational Agents 2nd Edition David L. Poole, Alan K. Mackworth - Solutions

13. Represent the electrical domain of previous chapters in ICL, so that it will run in AILog. The representation should include the probabilistic dependencies of Example 8.17 and the relations of Example 13.12.
12. Suppose you have a relational probabilistic model for movie prediction, which represents P(likes(P, M) ∣ age(P), genre(M))where age(P) and genre(M) are a priori independent.(a) What is the treewidth of the ground belief network (after pruning irrelevant variables)for querying age(Sam) given
11. For the representation of addition in Example 15.24, it was assumed that the observed Z-values would all be digits. Change the representation so that the observed values can be digits, a blank, or other. Give appropriate probabilities.
10. Suppose Boolean parameterized random variables young(Person) and cool(Item) are parents of Boolean buys(Person, Item). Suppose there are 3000 people and 200 items.(a) Draw this in plate notation.(b) How many random variables are in the grounding of this model?(c) How many numbers need to be
9. A simple modification for the gradient descent for collaborative filtering can be used to predict P(rating > threshold) for various values of threshold in {1, 2, 3, 4}. Modify the code so that it learns such a probability. [Hint: Make the prediction the sigmoid of the linear function as in
8. An alternative regularization for collaborative filtering is to minimize????⟨u, i, r⟩ ∈ D((^r (u, i) − r)2 + λ(ib[i]2 + ub[u]2 +???? p (ip[i, p]2 + up[u, p]2)))(a) How this differ from the regularization of Formula 15.1? [Hint: Compare the regularization for the items or users with few
7. Change the stochastic gradient descent algorithm of Figure 15.5 so it minimizes Formula 15.1, but regularizes after each example. Hint: You need to consider how many times each parameter is updated for one iteration through the data set and adjust the regularization parameter accordingly.
6. Give some concrete specialization operators that can be used for top-down inductive logic programming. They should be defined so that making progress can be evaluated myopically.Explain under what circumstances the operators will make progress.
5. Suppose that, in event calculus, there are two actions, Open and Close, and a relation opened that is initially, at time 0, false. Action Open makes opened true, and action Close makes opened false. Suppose that action Open occurs at time 5, and action Close occurs at time 10.(a) Represent this
4. In this exercise, you will investigate using event calculus for the robot delivery domain.(a) Represent the move action in the event calculus.(b) Represent each of the sequences of actions in Example 15.10 in the event calculus.(c) Show that event calculus can derive the appropriate goals from
3. AILog performs depth-bounded search. You will notice that the processing time for the previous questions was slow, and you required a depth bound that was close to the actual depth bound to make it work in a reasonable amount of time.In this exercise, estimate how long an iterative deepening
2. In this exercise, you will add a more complicated paint action than in the previous exercise.Suppose the object paint_can(Color) denotes a can of paint of color Color.Add the action paint(Obj, Color) that results in the object changing its color to Color.(Unlike in the previous question, the
1. Add to the situation calculus example (also available from the book web page) the ability to paint an object. In particular, add the predicate color(Obj, Col, Sit)that is true if object Obj has color Col in situation Sit.The parcel starts off blue. Thus, we have an axiom:color(parcel, blue,
• Plate models and the independent choice logic allow for the specification of probabilistic models before the individuals are known.
• Collaborative filtering can be used to make predictions about instances of relations from other instances by inventing hidden properties.
• Inductive logic programming can be used to learn relational models, even when the values of features are meaningless names.
• Event calculus allows for continuous and discrete time and axiomatizes what follows from the occurrence of events.
• The situation calculus represents time in terms of the action of an agent, using the init constant and the do function.
• Many of the representations in earlier chapters can be made relational.
• Relational representations are used when an agent requires models to be given or learned before it which individuals it will encounter.
17. In this question, you will write a meta-interpreter for parametrized logic programs. These are logic programs that can use constants in arithmetic expressions. The values for the constants are given as part of the input to the meta-interpreter.Assume that an environment is a list of terms of
16. Build an iterative deepening abductive reasoning system to find minimal consistent sets of assumables to imply a goal. This can be based on the depth-bounded meta-interpreter of Figure 14.12, and the delaying meta-interpreter of Figure 14.14 to collect assumptions. The depth bound should be
15. Write a meta-interpreter for definite clauses that does iterative deepening search. Make sure that it only returns one answer for each proof and that the system says no whenever the depthfirst searcher says no. This should be based on the depth-bounded meta-interpreter and the
14. Write a meta-interpreter that allows both how and why questions. In particular, it should allow the user to ask how questions about a goal that has been proved after a why question. Explain how such a program may be useful.
13. Write a program that takes in a tree produced from the meta-interpreter that builds proof trees as shown in Figure 14.13 and lets someone traverse the tree using how questions.
12. Extend the ask-the-user meta-interpreter from the previous question to allow for questions that ask for instances. The system could ask the user questions like “for which X is P(X) true?”, where the user can give an instance or tell the system there are no more instances. One feature that
11. Write a meta-interpreter that allows for asking the users yes-or-no questions. Make sure it does not ask questions to which it already knows the answer.
• askSite(URI, Q, Answer) is true when you ask the source URI a question Q, it gives the Answer that is one of {yes, no, unknown}. Note that although can_answer and reliability can be simple databases, askSite is a sophisticated program that accesses the web or asks a human a question.Write a
• reliability(URI, R) is true if R is some numerical measure of reliability of URI. You can assume that R is in the range [−100, 100], in which the higher number means that it is more reliable.
• can_answer(Q, URI) is true if the source given by URI can answer questions that unify with Q.
10. Write a meta-interpreter that can ask multiple sources for information. Suppose that each source is identified by a universal resource identifier (URI). Suppose you have the predicates
9. The program of Figure 14.14 allows duplicate delayed goals. Write a version of dprove that returns minimal sets of delayed goals, in their simplest form.
8. Consider two ways to modify the depth-bound meta-interpreter of Figure 14.12:(a) The bound is on number of instances of base-level atoms that appear in the proof.Why might this be better or worse than using the depth of the tree?(b) Allow different base-level atoms to incur different costs on
7. For the following, explain how each is categorized by the top-level ontology of Section 14.3.3:(a) your skin(b) the period at the end of the first sentence of this chapter(c) the excitement a child has before a vacation(d) the trip home from a vacation(e) a computer program(f) summer holidays(g)
6. A luxury hotel has multiple rooms to rent, each of which is comfortable and has a view. The hotel must also have more than one restaurant. There must be menu items for vegetarians and for meat eaters to eat in the restaurants.(a) Define a luxury hotel in OWL, based on this description. Make
5. Suppose a “beach resort” is a resort near a beach that the resort guests can use. The beach has to be near the sea or a lake, where swimming is permitted. A resort must have places to sleep and places to eat. Write a definition of beach resort in OWL.
4. Write an ontology for the objects that often appear on your desk that may be useful for a robot that is meant to tidy your desk. Think of the categories that (a) the robot can perceive and (b)should be distinguished for the task.
3. Sam has proposed that any n-ary relation P(X1, X2, X3, …, Xn) can be reexpressed as n − 1 binary relations, namely, P1(X1, X2) .P2(X2, X3) .P3(X3, X4) .⋮Pn − 1(Xn − 1, Xn) .Explain to Sam why this may not be such a good idea. What problems would arise if Sam tried to do this? Use an
2. A travel site has a database that represents information about hotels and feedback from users that uses the relations:hotel(Id, Name, City, Province, Country, Address)reported_clean(Hotel, RoomNumber, Cleanliness, day(Year, Month, Day))Show how the following facts can be represented using triple
1. There are many possible kinship relationships you could imagine like mother, father, great-aunt, second-cousin-twice-removed, and natural-paternal-uncle. Some of these can be defined in terms of the others, for example:brother(X, Y) ← father(X, Z) ∧ natural_paternal_uncle(Y, Z) .sister(X, Y)
• A meta-interpreter can be used to build a lightweight implementation of a knowledge-based system that can be customized to fit the requirements of the representation language.
• OWL ontologies are built from individuals, classes, and properties. A class is a set of real and potential individuals.
• Ontologies allow for semantic interoperability and knowledge sharing.
• Individual–property–value triples form a flexible, universal representation for relations.
17. Construct a knowledge base and a dictionary based on Figure 13.13 to answer geographical questions such as that given in Figure 1.2. For each query, either show how it can be answered or explain why it is difficult to answer given the tools presented in this chapter.
16. In this question, you are to write a definite clause knowledge base for the design of custom video presentations.Assume that the video is annotated using the relation segment(SegId, Duration, Covers), where SegId is an identifier for the segment. (In a real application this will be enough
15. The aim of this question is to get practice writing simple logic programs.(a) Write a relation remove(E, L, R) that is true if R is the list resulting from removing one instance of E from list L. The relation is false if E is not a member of L.(b) Give all of the answers to the following
14. Consider the following logic program:ap(emp, L, L) .ap(c(H, T), L, c(H, R)) ←ap(T, L, R) .adj(A, B, L) ←ap(F, c(A, c(B, E)), L) .(a) Give a top-down derivation (including all substitutions) for one answer to the query???????????? adj(b, Y, c(a, c(b, c(b, c(a, emp))))) .(b) Are there any
13. Consider the following logic program:rd(cons(H, cons(H, T)), T) .rd(cons(H, T), cons(H, R)) ←rd(T, R) .Give a top-down derivation, showing all substitutions for the query???????????? rd(cons(a, cons(cons(a, X), cons(B, cons(c, Z)))), W) .What is the answer corresponding to this derivation?Is
12. Consider the following logic program:f(empty, X, X) .f(cons(X, Y), W, Z) ←f(Y, W, cons(X, Z)) .Give each top-down derivation, showing substitutions (as in Example 13.32) for the query???????????? f(cons(a, cons(b, cons(c, empty))), L, empty) .What are all of the answers?
11. List all of the ground atomic logical consequences of the following knowledge base:q(Y) ← s(Y, Z) ∧ r(Z) .p(X) ← q(f(X)) .s(f(a),b) .s(f(b),b) .s(c,b) .r(b) .
10. For each of the following pairs of atoms, either give a most general unifier or explain why one does not exist:(a) p(X, Y,a, b, W)p(E,c, F, G, F)(b) p(X, Y, Y)p(E, E, F)(c) p(Y,a, b, Y)p(c, F, G, F)(d) ap(F0, c(b, c(B0, L0)), c(a, c(b, c(b, c(a, emp)))))ap(c(H1, T1), L1, c(H1, R1))
9. Give a most general unifier of the following pairs of expressions:(a) p(f(X), g(g(b))) and p(Z, g(Y))(b) g(f(X), r(X), t) and g(W, r(Q), Q)(c) bar(val(X, bb), Z) and bar(P, P)
8. What is the result of the following applications of substitutions?(a) f(A, X, Y, X, Y){A/X, Z/b, Y/c} .(b) yes(F, L) ← append(F, c(L, nil), c(l, c(i, c(s, c(t, nil))))){F/c(l, X1), Y1/c(L, nil), A1/l, Z1/c(i, c(s, c(t, nil)))} .(c) append(c(A1, X1), Y1, c(A1, Z1)) ← append(X1, Y1, Z1){F/c(l,
7. Consider the following knowledge base:has_access(X, library) ← student(X) .has_access(X, library) ← faculty(X) .has_access(X, library) ← has_access(Y, library) ∧ parent(Y, X) .has_access(X, office) ← has_keys(X) .faculty(diane) .faculty(ming) .student(william) .student(mary)
6. In a manner similar to Example 13.23, show derivations of the following queries:(a) ???????????? two_doors_east(r107, R) .(b) ???????????? next_door(R, r107) .(c) ???????????? west(R, r107) .(d) ???????????? west(r107, R) .Give all answers for each query.
5. In Example 13.23, we always selected the leftmost conjunct to resolve on. Is there a selection rule (a selection of which conjunct in the query to resolve against) that would have resulted in only one choice for this example? Give a general rule that – for this example, at least – results in
4. In Example 13.23, the algorithm fortuitously chose imm_west(r109, r111) as the clause to resolve against. What would have happened if another clause had been chosen? Show the sequence of resolutions that arise, and either show a different answer or give a generalized answer clause that cannot
3. Consider the following knowledge base:r(a) .r(e) .p(c) .q(b) .s(a,b) .s(d,b) .s(e,d) .p(X) ← q(X) ∧ r(X) .q(X) ← s(X, Y) ∧ q(Y) .Show the set of ground atomic consequences derivable from this knowledge base. Assume that a bottom-up proof procedure is used and that at each iteration the
2. Consider the language that contains the constant symbolsa, b, and c; the predicate symbols p and q; and no function symbols. We have the following knowledge bases built from this language:KB1 = { p(a) }KB2 = { p(X) ← q(X) }KB3 = { p(X) ← q(X), p(a), q(b) } .Now consider possible
1. Consider a domain with two individuals (✂ and ☎), two predicate symbols (p and q), and three constants (a,b, and c). The knowledge base KB is defined by p(X) ← q(X) .q(a) .(a) Give one interpretation that is a model of KB.(b) Give one interpretation that is not a model of KB.(c) How many
• Clark’s completion can be used to define the semantics of negation as failure under the complete knowledge assumption.
• Equality between terms means that the terms denote the same individual.
• It is possible to use definite clauses to represent natural language grammars.
• Function symbols are used to denote a possibly infinite set of individuals described in terms of other individuals. Function symbols can be used to build data structures.
• Substitutions are used to make instances of atoms and rules. Unification makes atoms identical for use in proofs.
• Datalog is a logical language with constants, universally quantified variables, relations, and rules.
• In domains characterized by individuals and relations, constants denoting individuals and predicate symbols denoting relations can be reasoned with to determine what is true in the domain.
10. Consider the policy improvement algorithm. At equilibrium the values of the most-preferred actions should be equal. Propose, implement and evaluate an algorithm where the policy does not change very much when the values of the most-preferred actions are close. [Hint: Consider having the
9. In Example 12.6, some of the features are perfectly correlated (e.g., F6 and F7). Does having such correlated features affect what functions are able to be represented? Does it help or hurt the speed at which learning occurs? Test this empirically on some examples.
8. In SARSA with linear function approximation, using linear regression to minimize r + γQw¯¯¯(s′, a′) − Qw¯¯¯(s, a), gives a different algorithm than Figure 12.7. Explain what you get and why what is described in the text may be preferable (or not).
7. Included the features for the grid game of Example 12.6, are features that are the x-distance to the current treasure and features that are the y-distance to the current treasure. Chris thought that these were not useful as they do not depend on the action. Do these features make a difference?
6. The model-based reinforcement learner allows for a different form of optimism in the face of uncertainty. The algorithm can be started with each state having a transition to a “nirvana” state, which has very high Q-value (but which will never be reached in practice, and so the probability
5. Consider four different ways to derive the value of αk from k in Q-learning (note that for Qlearning with varying αk, there must be a different count k for each state–action pair).(a) Let αk = 1/k.(b) Let αk = 10/(9 + k).(c) Let αk = 0.1.(d) Let αk = 0.1 for the first 10,000 steps, αk =
4. For the following reinforcement learning algorithms:(a) Q-learning with fixed α and 80% exploitation.(b) Q-learning with fixed αk = 1/k and 80% exploitation.(c) Q-learning with αk = 1/k and 100% exploitation.(d) SARSA learning with αk = 1/k and 80% exploitation.(e) SARSA learning with αk =
3. Compare the different parameter settings for the game of Example 12.2. In particular compare the following situations(a) α varies, and the Q-values are initialized to 0.0(b) α varies, and the Q-values are initialized to 5.0(c) α is fixed to 0.1, and the Q-values are initialized to 0.0(d) α
2. For the plot of the total reward as a function of time as in Figure 12.4, the minimum and zero crossing are only meaningful statistics when balancing positive and negative rewards is reasonable behavior. Suggest what should replace these statistics when zero reward is not an appropriate
1. Explain how Q-learning fits in with the agent architecture of Section 2.2.1. Suppose that the Qlearning agent has discount factor γ, a step size of α, and is carrying out an ϵ-greedy exploration strategy.(a) What are the components of the belief state of the Q-learning agent?(b) What are the
• Model-based reinforcement learning separates learning the dynamics and reward models from the decision-theoretic planning of what to do given the models.
• Off-policy learning, such as Q-learning, learns the value of the optimal policy. On-policy learning, such as SARSA, learns the value of the policy the agent is actually carrying out (which includes the exploration).
• In reinforcement learning, an agent should trade off exploiting its knowledge and exploring to improve its knowledge.
• A Markov decision process is an appropriate formalism for reinforcement learning. A common method is to learn an estimate of the value of doing each action in a state, as represented by the Q(S, A) function.
7. Consider the sequential prisoner’s dilemma.(a) Suppose the agents play for a fixed number of times (say three times). Give two equilibria if there are two or more, otherwise give the unique equilibrium and explain why there is only one. Hint: Consider the last time first.(b) Suppose there is a
6. Answer the same questions as in the previous exercise for the following games:(a)D E F A2, 1110, 103, 12 B 5, 7 12, 1 6, 5 C 6, 5 13, 2 4, 6 (b)D E F A80, 130 20, 10 130, 80 B130, 80 30, 20 80, 130 C 20, 10 100, 100 30, 20
5. Consider the following normal-form game where the row player can choose action A, B or C and the column player could choose action D, E, or F:D E F A 40, 40 120, 10 60, 30 B 30, 60 110, 60 90, 90 C30, 110100, 10070, 120 where the pairs give the value of the outcome for the row player followed by
4. In Example 11.12, what is the Nash equilibrium with randomized strategies? What is the expected value for each agent in this equilibrium?
3. Which of the following games in normal form have a Nash equilibrium made up of pure strategies? For those that do, specify the pure strategy Nash equilibria. For those that do not, explain how you know there is no pure strategy Nash equilibrium.
2. For the hawk–dove game of Example 11.11, where D > 0 and R > 0, each agent is trying to maximize its utility. Is there a Nash equilibrium with a randomized strategy? What are the probabilities? What is the expected payoff to each agent? (These should be expressed as functions of R and D). Show
1. Consider the game of Tic-Tac-Toe (also called Noughts and Crosses), which is played by two players, an “X” player and an “O” player who alternate putting their symbol in a blank space on a 3 × 3 game-board. A player’s goal is to win by placing three symbols in a row, column, or
• By introducing payments, it is possible to design a mechanism that is dominant-strategy truthful and economically efficient.
• A Nash equilibrium is a strategy profile for each agent such that no agent can increase its utility by unilaterally deviating from the strategy profile.
• In partially observable domains, sometimes it is optimal to act stochastically.
• Perfect-information games can be solved by backing up values in game trees or searching the game tree using minimax with α-β pruning.
• A multiagent decision network models probabilistic dependency and information availability.
• The extensive form of a game models agents’ actions and information through time in terms of game trees.
• The strategic form or normal form of a game specifies the expected outcome given controllers for each agent.
• A multiagent system consists of multiple agents who act autonomously and have their own utility over outcomes. The outcomes depend on the actions of all the agents. Agents can compete, cooperate, coordinate, communicate, and negotiate.
7. As outlined in Example 10.7, define a code for describing decision trees. Make sure that each code corresponds to a decision tree (for every sufficiently long sequence of bits, the initial segment of the sequence will describe a unique decision tree), and each decision tree has a code.How does
6. To initialize the EM algorithm in Figure 10.6 consider two alternatives:(a) allow P to return a random distribution the first time through the loop(b) initialize cc and fc to random values By running the algorithm on some data sets, determine which, if any, of these alternatives is better in
5. Suppose the k-means algorithm is run for an increasing sequence of values for k, and that it is run for a number of times for each k to find the assignment with a global minimum error. Is it possible that a number of values of k exist for which the error plateaus and then has a large improvement

Showing 2400 - 2500 of 4756

First
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
Last

Step by Step Answers

Services

Sitemap
Fun
Definitions
Become Tutor
Used Textbooks
Study Help Categories
Recent Questions
Expert Questions
Campus Wear
Sell Your Books

Company Info

Security
Copyrights
Privacy Policy
Terms & Conditions
SolutionInn Fee
Scholarship
Online Quiz
Give Feedback, Get Rewards

Get In Touch

About Us
Contact Us
Career
Jobs
FAQ
Student Discount
Campus Ambassador

Secure Payment

Download Our App

© 2026 SolutionInn. All Rights Reserved