AI Final

Assume P(a)=0.3, P(b|a)=0.2, and P(c|a,b)=0.1. What is the value of P(a,b,c) ? 0 Cannot calculate without additional information 0.1 0.006

0.006

What is the expected utility of buying the book? 1620 1300 0.86 0.65

1620

If a search algorithm has an effective branching factor of 1.5 and the solution depth is 5, then approximately how many nodes does it generate during the search? 11 20 2 7.5

20

If the minimax algorithm with alpha-beta pruning is used to find the optimal strategy for this game, then how many times will a pruning be performed? 3 4 2 1 0

4

If there are four binary variables in a probabilistic model, but two of them are independent of the rest, then what is the minimum number of values required to represent the model? 4 6 5 8

5

What is the expectiminimax value of the following game tree? 7/3 1 5 6

6

Assume the following confusion matrix: What is the value of the F1 measure? 0 3/4 6/7 1

6/7

Given the full joint distribution shown above, calculate P(Cavity). 0.8 <0.4, 0.6> 0.2 <0.2, 0.8>

<0.2, 0.8>

Given the full joint distribution shown above, calculate P(Cavity|toothache ∨ catch). <0.4615, 0.5384> <0.192, 0.224> <0.871, 0.129> <0.108, 0.016>

<0.4615, 0.5384>

The following is an example of what type of structure? Belief graph Update tree AND-AND graph AND-OR tree

AND-OR tree

Which of the following property is NOT required by database semantics: All are required Closed-world assumption Domain closure Unique-names assumption

All are required

Assume you have two choices: parachute out of an airplane (A) or attempt to climb a mountain (M). If you choose A, then there could be two outcomes: you successfully land (with probability of 0.99) or you die. If you choose M, you can successfully reach the summit and come down (with probability of 0.99), or you could die along the way. Successfully parachuting out of the airplane yields a utility of 100, while successfully reaching the summit of Mt. Everest and coming down yields an utility of 200. Which choice has the higher expected utility? There is not enough information in the problem to make the determination. Since dying is infinitely worse than anything else, the expected utility of both actions is the same. Parachuting out of an airplane (A) Attempting to climb a mountain (M)

Attempting to climb a mountain (M)

The Prolog language is based on which of the following? Model checking Forward chaining Backward chaining Resolution

Backward chaining

Which of the following search algorithms is NOT optimal? Uniform-cost search Iterative deepening search Depth-first search Breadth-first search

DFS

Assume you designed an email spam filter that detects whether an incoming email should be put into a spam folder (positive case). If the filter has high precision, but low recall, then what will be the effect? Emails that end up in the spam folder will very likely be actual spam, but not many actual spam emails get detected. Many non-spam emails will end up in the spam folder. Most spam emails will end up in the spam folder. Emails that end up in the spam folder will unlikely be actual spam, but many actual spam emails get detected.

Emails that end up in the spam folder will very likely be actual spam, but not many actual spam emails get detected.

Which English statement is closest to the meaning of the following logic statement? ∀x∃y Dog(x)⇒Day(y)∧Has(x,y). For every dog, there is a day. There is a day in which all dogs have a day. Every dog has its day. All days have dogs.

Every dog has its day.

Which English statement is closest to the meaning of the following logic statement? ∀x∃y Dog(x)⇒Day(y)∧Has(x,y). There is a day in which all dogs have a day. For every dog, there is a day. Every dog has its day. All days have dogs.

Every dog has its day.

A Skolem constant is used in which of the following? Universal generalization Existential instantiation Universal instantiation Existential generalization

Existential instantiation

The problem of not knowing whether to make use of learned knowledge or continue learning in an online environment is called the: Exploration vs. exploitation problem Conformant problem Random walk problem Competitive ratio problem

Exploration vs. exploitation problem`

A single perceptron has a larger hypothesis function space than a decision tree. (T or F)

F

Assume you can play a $1,000,000,000,000 lottery where the expected value of the prize is $1.25. The ticket costs $2. Buying the lottery ticket would always be irrational. (T or F)

F

The probability of a disjunction of two events is the sum of their individual probabilities. (T or F)

F

An admissible heuristic is always consistent. (T or F)

False

Bidirectional search has both better time and space complexity than breadth-first search. Both are complete and optimal. Therefore, bidirectional search should always be used instead of breadth-first search. (T or F)

False

Conformant problems are search problems that are unsolvable. (T or F)

False

Omniscience is a sufficient condition for rationality. (T or F)

False

The purpose of a fitness function in a genetic algorithm is to determine the crossover point. (T or F)

False

After your yearly checkup, the doctor has bad news and good news. The bad news is that you tested positive for a serious disease and that the test is 99% accurate (i.e. the probability of testing positive, when you do have the disease is 0.99 as is the probability of testing negative when you don't have the disease). The good news is that this is a rare disease, striking only 1 in 10,000 people of your age. Which of the following statements accurately describes the consequence that the disease is rare? It means the test is not actually accurate, so there is a low chance of getting the disease. Few people get it (prior probability is low), so by Bayes' theorem, the posterior probability is low regardless of the test result. It means that the prior probability is high, so by Bayes' theorem, the posterior probability is also high. It's only likely to affect the people that took the test, so the posterior probability is high.

Few people get it (prior probability is low), so by Bayes' theorem, the posterior probability is low regardless of the test result

If you perform simulated annealing with the temperature parameter kept at zero at all times, then the algorithm is equivalent to which of the following (disregard the termination check, i.e. "if T=0 then return current") ? First-choice hill climbing Random walk Beam search Stochastic hill climbing

First-choice hill climbing

Which of the following properties best characterize the task environment of "Photographing tennis games. Specifically, choosing the moment for capturing a photo" ? Discrete Dynamic Fully Observable Single-Agent Partially Observable Static Multi-agent Continuous

Fully, Single, Static, Continuous

Which of the following does NOT deal with uncertainty of events? Fuzzy logic Fudge factors Probability theory Default logic

Fuzzy logic

Which English statement is closest to the meaning of the following logic statement? ∀x,y,l Person(x)∧Person(y)∧Language(l)∧Speaks(x,l)∧Speaks(y,l)⇒Understands(x,y)∧Understands(y,x). If there is a language that two people can understand, then everyone will understand them. All people understand the same language For all languages spoken by people, only two people can understand them. If two people speak the same language, they will understand each other

If two people speak the same language, they will understand each other

Which of the following are needed for a complete search problem formulation? Search Algorithm Initial State Path Cost Transition Model Actions Goal Test

Initial State Path Cost Transition Model Actions Goal Test

What is a Turing Test? It is a test used to determine computer's intelligence. It involves having a human decide whether it is communicating with a computer or a human being. It is a test designed to fool computer artificial intelligence. It is a test of human intelligence. It involves having a computer decide whether it is communicating with a computer or a human being. It is a thought experiment used to think about whether an inanimate object can posses intelligence.

It is a test used to determine computer's intelligence. It involves having a human decide whether it is communicating with a computer or a human being.

What is the order in which nodes are examined when executing the A* search algorithm from Lugoj to Bucharest? (map and heuristic cost to destinations are given below) Lugoj, Timisoara, Mehadia, Drobeta, Craiova, Pitesti, Bucharest Lugoj, Mehadia, Drobeta, Craiova, Pitesti, Bucharest Lugoj, Mehadia, Drobeta, Craiova, Timisoara, Pitesti, Bucharest Lugoj, Mehadia, Drobeta, Craiova, Rimnicu Vilcea, Timisoara, Pitesti, Bucharest

Lugoj, Mehadia, Drobeta, Craiova, Timisoara, Pitesti, Bucharest

Which of the following heuristics picks variables with fewest legal values first? Minimum-remaining-values heuristic Degree heuristic Least-constraining-value heuristic Min-conflicts heuristic

Minimum-remaining-values heuristic

Assume you have the following logic statement: (Smoke⇒Fire) ⇒ (¬Smoke⇒¬Fire) This statement can be characterized as: Cannot be determined Valid Unsatisfiable Neither valid or unsatisfiable

Neither valid or unsatisfiable

The idea that choosing a simpler model provides better machine learning performance is supported by: Ockham's Razor Decision trees No free lunch theorem Overfitting

Ockham's Razor

Which of the following properties best characterize the task environment of "Playing Soccer" ? Partially Observable Single-agent Episodic Deterministic Dynamic Sequential Multi-agent Fully Observable Static Continuos Stochastic Discrete

Partially, Dynamic, Sequential, Multi, Continuos, Stochastic

What does PEAS stand for?

Performance measure, Environment, Actuators, Sensors

If a person works hard, then she will pass the AI course. Jennifer works hard. Therefore, Jennifer will pass the AI course. Modus Ponens Modus Tollens De Morgan's Law Resolution

Ponens

Assume you have the data set given below, which provides hypothetical examples of instances when people did or did not get hired for a job. It consists of three categorical attributes and a label that indicates "hired" or "not hired". Which attribute would be chosen as the root of the decision tree if the tree is induced using in a greedy manner using information gain as the criterion? REFOP It would be randomly chosen QUAL EXP

REFOP

If you perform simulated annealing with the temperature parameter kept at infinity at all times, then the algorithm is equivalent to which of the following? First-choice hill climbing Random walk Beam search Stochastic hill climbing

Random walk

The machine learning task that aims to learn the safe following distance given car crash data is a: Regression task Classification task Reinforcement learning task Clustering task

Regression task

Assume you use a greedy search to choose the shortest path from S to D: S, A, B, D S, A, C, D S, A, G, C, G, D, G S, G

S, A, B, D

The set of all atomic events is called the: random variable probability model event sample space

Sample Space

If a model estimates the probability of a state using the values of the past two states, it is a First-order Markov process Stationary process Utility model Second-order Markov process

Second-order Markov process

A CSP problem that is made strongly n-consistent can be solved in polynomial time complexity. (T or F)

T

Bayesian networks cannot contain cycles. (T or F)

T

Decision networks are Bayesian networks with action and utility nodes. (T or F)

T

Every search problem with a partially observable environment can be solved by transforming it into one that is fully observable. (T or F)

T

If agent's choices only depend on deterministic prizes, then the actual assigned utility values do not matter as long as they preserve the preference order. (T or F)

T

The resolution algorithm can only be used if the KB is in Conjunctive Normal Form. (T or F)

T

What happens when the value k in the k'th nearest neighbor algorithm is increased? The learning is more generalizible, but there will be more variance in the errors. The algorithm will take less data points into account when making classification decisions The variance in the errors decreases, but learning because more specific. The algorithm will be more likely to overfit.

The learning is more generalizible, but there will be more variance in the errors.

Assume you have a vacuum world with the following belief space: The goal of the vacuum world is to get to a state with no dirt in it (shown by dots in the squares of the states). If the initial state is the middle one (belief state of all physical states), then what is the maximum number of moves needed before the agent gets to the goal? 1 2 4 This problem has no solution

This problem has no solution

What is the purpose of an evaluation function when using a minimax search? To cutoff search at some point and evaluate the likely payoff from that point To evaluate the quality of the adversary. To evaluate whether a state satisfies the goal test. To determine the number of moves necessary to win the game.

To cutoff search at some point and evaluate the likely payoff from that point

A reflex agent with a randomized agent function can outperform a simple reflex agent. (T or F)

True

A structured environment representation is generally more complex and difficult to learn than a factored representation. (T or F)

True

Every First-Order Logic statement can be propositionalized so as to preserve entailment. (T or F)

True

In a Bayesian Network with only binary variables, if each variable has no more than 5 parents, then the space complexity is linear in the number of variables. (T or F)

True

If someone is a king then they have a crown. John is a king, so John has a crown. This is an example of Existential instantiation Existential generalization Universal generalization Universal instantiation

Universal instantiation

Assume you have the following logic statement: (Smoke ⇒Fire)⇒((Smoke ∧Heat)⇒Fire) This statement can be characterized as: Unsatisfiable Valid Cannot be determined Neither valid or unsatisfiable

Valid

Assume you have the following logic statement: Big ∨ Dumb ∨ (Big ⇒ Dumb) This statement can be characterized as: Neither valid or unsatisfiable Cannot be determined Unsatisfiable Valid

Valid

Assume you have the following logic statement: Smoke => Smoke This statement can be characterized as:

Valid

The inference task that estimates the probability of the current belief state of the agent is called filtering prediction smoothing most likely explanation task

filtering

Which of the following is NOT an example of an ensemble machine learning method? AdaBoost k'th nearest neighbor Gradient Boosted Decision Trees Random Forest

k'th nearest neighbor

The inference task that estimates the probability of future states is called most likely explanation task prediction filtering smoothing

prediction

If an agent prefers a fixed sum of money to a lottery with the same expected monetary value, then the agent is rational omniscient risk-averse risk-pron

risk-averse

If an agent prefers a fixed sum of money to a lottery with the same expected monetary value, then the agent is risk-prone risk-averse rational omniscient

risk-averse

The inference task that estimates the probability of past states is called: prediction filtering most likely explanation task smoothing

smoothing

A logical expression that refers to an object in first-order predicate calculus is called a term constant object predicate

term

What is the unification of the following statements? p is Likes(John, x) q is Likes(y, Mary) {x/Mary, y/John} {x/Mary, y/Mary} Cannot unify {x/John, y/x}

{x/Mary, y/John}

Which of the following first logic statements is equivalent to "For every recursive solution, there is an iterative solution." ∃x ∃y [RecursiveSolution(x)⇒IterativeSolution(y)] ∃x ∀y [RecursiveSolution(x)⇒IterativeSolution(y)] ∀x ∃y [RecursiveSolution(x)⇐IterativeSolution(y)] ∀x ∃y [RecursiveSolution(x)⇒IterativeSolution(y)]

∀x ∃y [RecursiveSolution(x)⇒IterativeSolution(y)]