IAI Questions

Last updated about 3 years ago

41 questions

1

Library

IAI Questions

Last updated about 3 years ago

41 questions

1

Other Answer Choices:

1

Draggable item		Corresponding Item

1

Other Answer Choices:

d is finite

1

Other Answer Choices:

1

Consider map coloring problem, which is CSP. Here, adjacent regions must have different colors. We have 3 regions A, B, C, that are all adjacent to each other.
A is not equal to B is __________ constraint. The __________ of the constraint is (A,B). We can add __________ such as red is better than green.

1

Backtracking search is __________ algorithm. It is __________ with constraint checking. It checks constraints __________. It backtracks when __________ .

1

Draggable item		Corresponding Item

1

Draggable item		Corresponding Item

1

Other Answer Choices:

1

________ add new sentences to knowledge base. Agent can use inference to deduce new facts from _________ ed facts.

Other Answer Choices:

TELL

ASK

1

________________ commitment of propositional language is fact. ____________________ commitment is T/F/unknown.

Other Answer Choices:

Epistemological

Ontological

1

Below sentence explains the syntax of FOL. ^ ________________ . ∀ is ________________ . 

Other Answer Choices:

connectives

quantifiers

1

Draggable item		Corresponding Item

1

__________ allows incremental updating of beliefs
as more evidence is gathered.

1

Bayesian networks describe __________ distributions using __________ distributions.

Other Answer Choices:

local

joint

1

Draggable item		Corresponding Item

1

Observing the cause ________________  the path between two effects of same cause. Observing the cause ______________ the path between two causes of same effect.

Other Answer Choices:

inactivates

activates

What is wrong about discounting reward?

Later reward probably have higher utility than sooner reward

Discounting reward helps algorithm converge

Discounting factor is between 0 and 1

Below explains some properties of RL. What is incorrectly connected?

Exploration - risk of getting negative reward

Markov - action outcomes depend only on the current state

Policy - single plan that can achieve goal states

Exploitation - makes the decision that maximizes the long term sum of rewards

What is wrong about optimal value functions?

Optimal policy is a consequence of optimal value function

We can solve it in cubic time by standard linear algebra solvers.

Optimal value functions can be decomposed.

They specify the best possible performance over all policies.

We stop the value iteration when:

Maximum change in utility of state is higher than threshold

Maximum change in utility of state is lower than threshold

After N iterations

Minimum change in utility of state is higher than threshold

 The time complexity of value iteration is ________   per iteration, where policy evaluation is ________  per iteration.

Other Answer Choices:

What is the advantage of policy iteration over value iteration?

It gains efficiency in policy improvement phase.

The optimal policy can converge before the values converge.

Does not calculate and converges value function 

Policy is extracted using one-step look ahead.

Match the algorithms with appropriate target.

Draggable item		Corresponding Item
Q-learning
TD(0)
MC
SARSA

Categorize algorithms.

Value iteration
MC
TD
DQN

Model-free
Model-based

Categorize algorithms.

DQN
Q-learning
SARSA
TD

On-policy
Off-policy

MC
TD
DP
SARSA

Bootstrapping
Sampling

What is wrong about DQN? Choose all.

Q prediction network - determine the next action

Experience replay - inefficient since passes the same data multiple times

Q' target network - move target closer to prediction

Q prediction network - chase a stationary target

Experience replay - removes data correlation by batch sampling

What is wrong about the below explanations about the representation of search problems?

Each node in search tree is a path in state space graph.

Each state occurs only once in state space graph.

Connected node represent action results.

The initial state is root node of search tree.

What is true about search problem and algorithms? Choose all.

Time complexity is the number of nodes generated during search.

Space complexity is the total number of nodes stored in memory during search.

Reflex agents aims to reach the goal.

We should maintain the unexplored nodes for efficiency.

What is the data structure of frontier by the algorithm?

Tree search
Graph search
DFS
UCS
A*
Best first search

Closed list
Queue
Priority queue

DLS is complete if ________ and optimal if ________  where l denotes depth limit, m denotes max depth, and d denotes the depth of the shallowest goal.

Other Answer Choices:

m is finite

d is finite

(Row) expands the (Column) unexplored node in the frontier. Choose the right

The space complexity of frontier of BFS is ________ and explored list ________ . The space complexity of frontier of DFS is ________ .

Other Answer Choices:

When is inappropriate to use the bi-directional search?

There are multiple goal states.

Predecessor function is infeasible to compute.

The max depth of the search is infinite.

Successor function is efficiently computable.

What is wrong about DLS? Choose all.

Time complexity is same as BFS.

Space complexity is same as DFS.

Combine DFS’s space advantage with BFS’s completeness.

Preferred when maximum depth of a solution is known.

What is true about UCS? Choose all.

It replace the node in frontier with the node with lower cost.

It explores decreasing cost contours.

It explores options in every direction.

It uses goal information.

It may result in sub-optimal path.

What is true about A* search?

The first goal node opened may be sub-optimal, but finds optimal path if heuristic is admissible and consistent.

It avoids expanding paths that are already expensive.

It reduce space complexity from exponential to quadratic by using a good heuristic. 

We can weight the heuristic to find optimal solution faster.

What is wrong about relaxed problems?

Cost of the optimal solution in the relaxed problem is an admissible heuristic in the original problem.

The removal of restrictions removes edges from state space graph.

Optimal solution in the original problem is also a solution for the relaxed problem.

Relaxed problems have fewer restrictions than original problems.

What is wrong about approaches for obtaining heuristics?

We can combine admissible heuristics.

We can compare heuristics and choose heuristic with higher branching factor.

We can learn heuristic by identifying which features are predictive of path cost.

We can derive heuristic from the cost of a subproblem.

Categorize local search methods.

Hill-climbing
Stochastic hill-climbing
Simulated annealing
Local beam search

Select highest valued neighbor
Select neighbor that produces an improvement
Accept bad moves

Consider map coloring problem, which is CSP. Here, adjacent regions must have different colors. We have 3 regions A, B, C, that are all adjacent to each other.
A is not equal to B is __________ constraint. The __________ of the constraint is (A,B). We can add __________ such as red is better than green.

Backtracking search is __________ algorithm. It is __________ with constraint checking. It checks constraints __________. It backtracks when __________ .

Match the explanation of approaches of ordering variable in CSP.

Draggable item		Corresponding Item
Degree heuristic		The variable with the fewest remaining legal values in its domain
Most constrained variable		The variable involved in the largest number of constraints on other unassigned variables.

Match the explanation of approaches of filtering variable in CSP.

Draggable item		Corresponding Item
Arc consistency		Propagation from assigned variable to unassigned variable
Both		Examines further implications, not only 1-step
Forward checking		Detects failures earlier

The time complexity of AC-3 is O(?) because there are ________ edges, take________ time for consistency enforcing, and ________ for arc insertions.

Other Answer Choices:

________ add new sentences to knowledge base. Agent can use inference to deduce new facts from _________ ed facts.

Other Answer Choices:

TELL

ASK

________________ commitment of propositional language is fact. ____________________ commitment is T/F/unknown.

Other Answer Choices:

Epistemological

Ontological

What can be inferred from below?

Below sentence explains the syntax of FOL. ^ ________________ . ∀ is ________________ . 

Other Answer Choices:

connectives

quantifiers

Match the explanation of well-known problems of planning.

Draggable item		Corresponding Item
Ramification problem		Representing all things that stay the same
Frame problem		Defining conditions for an action to succeed
Qualification problem		Representing implicit consequences of actions

What is true about linear planning?

It interleaves the goal achievement.

It can lead to Sussman anomaly. 

It is efficient since it does not maintains goal stack.

It is useful when multiple goals interact, since it increases efficiency by separating goals.

Consider PDDL. What is included in what?

Actions
Initial state
Goal
Types
Predicates
Objects
Constants

Domain definition
Problem definition

What is wrong about Bayes Rule?

We can build conditional from its reverse.

Posterior is proportional to likelihood time prior.

Likelihood represents the agent's beliefs given the evidence. 

Posterior and likelihood are a conditional.

__________ allows incremental updating of beliefs
as more evidence is gathered.

Bayesian networks describe __________ distributions using __________ distributions.

Other Answer Choices:

local

joint

Match the type of causal chain with the definition of joint distribution.

Draggable item		Corresponding Item
Common cause
Common effect
Casual chain

Observing the cause ________________  the path between two effects of same cause. Observing the cause ______________ the path between two causes of same effect.

Other Answer Choices:

inactivates

activates

What is wrong about discounting reward?

Later reward probably have higher utility than sooner reward

Discounting reward helps algorithm converge

Discounting factor is between 0 and 1

Below explains some properties of RL. What is incorrectly connected?

Exploration - risk of getting negative reward

Markov - action outcomes depend only on the current state

Policy - single plan that can achieve goal states

Exploitation - makes the decision that maximizes the long term sum of rewards

What is wrong about optimal value functions?

Optimal policy is a consequence of optimal value function

We can solve it in cubic time by standard linear algebra solvers.

Optimal value functions can be decomposed.

They specify the best possible performance over all policies.

We stop the value iteration when:

Maximum change in utility of state is higher than threshold

Maximum change in utility of state is lower than threshold

After N iterations

Minimum change in utility of state is higher than threshold

 The time complexity of value iteration is ________   per iteration, where policy evaluation is ________  per iteration.

What is the advantage of policy iteration over value iteration?

It gains efficiency in policy improvement phase.

The optimal policy can converge before the values converge.

Does not calculate and converges value function 

Policy is extracted using one-step look ahead.

Match the algorithms with appropriate target.

Q-learning

TD(0)

MC

SARSA

Categorize algorithms.

Value iteration

MC

TD

DQN

Model-free

Model-based

Categorize algorithms.

DQN

Q-learning

SARSA

TD

On-policy

Off-policy

MC

TD

DP

SARSA

Bootstrapping

Sampling

What is wrong about DQN? Choose all.

Q prediction network - determine the next action

Experience replay - inefficient since passes the same data multiple times

Q' target network - move target closer to prediction

Q prediction network - chase a stationary target

Experience replay - removes data correlation by batch sampling

What is wrong about the below explanations about the representation of search problems?

Each node in search tree is a path in state space graph.

Each state occurs only once in state space graph.

Connected node represent action results.

The initial state is root node of search tree.

What is true about search problem and algorithms? Choose all.

Time complexity is the number of nodes generated during search.

Space complexity is the total number of nodes stored in memory during search.

Reflex agents aims to reach the goal.

We should maintain the unexplored nodes for efficiency.

What is the data structure of frontier by the algorithm?

Tree search

Graph search

DFS

UCS

A*

Best first search

Closed list

Queue

Priority queue

DLS is complete if ________ and optimal if ________  where l denotes depth limit, m denotes max depth, and d denotes the depth of the shallowest goal.

m is finite

(Row) expands the (Column) unexplored node in the frontier. Choose the right 

Deepest

Shallowest

Cheapest

BFS

DFS

DLS

UCS

IDS

The space complexity of frontier of BFS is ________ and explored list ________ . The space complexity of frontier of DFS is ________ .

When is inappropriate to use the bi-directional search?

There are multiple goal states.

Predecessor function is infeasible to compute.

The max depth of the search is infinite.

Successor function is efficiently computable.

What is wrong about DLS? Choose all.

Time complexity is same as BFS.

Space complexity is same as DFS.

Combine DFS’s space advantage with BFS’s completeness.

Preferred when maximum depth of a solution is known.

What is true about UCS? Choose all.

It replace the node in frontier with the node with lower cost.

It explores decreasing cost contours.

It explores options in every direction.

It uses goal information.

It may result in sub-optimal path.

What is true about A* search?

The first goal node opened may be sub-optimal, but finds optimal path if heuristic is admissible and consistent.

It avoids expanding paths that are already expensive.

It reduce space complexity from exponential to quadratic by using a good heuristic. 

We can weight the heuristic to find optimal solution faster.

What is wrong about relaxed problems?

Cost of the optimal solution in the relaxed problem is an admissible heuristic in the original problem.

The removal of restrictions removes edges from state space graph.

Optimal solution in the original problem is also a solution for the relaxed problem.

Relaxed problems have fewer restrictions than original problems.

What is wrong about approaches for obtaining heuristics?

We can combine admissible heuristics.

We can compare heuristics and choose heuristic with higher branching factor.

We can learn heuristic by identifying which features are predictive of path cost.

We can derive heuristic from the cost of a subproblem.

Categorize local search methods.

Hill-climbing

Stochastic hill-climbing

Simulated annealing

Local beam search

Select highest valued neighbor

Select neighbor that produces an improvement

Accept bad moves

Match the explanation of approaches of ordering variable in CSP.

Degree heuristic

The variable with the fewest remaining legal values in its domain

Most constrained variable

The variable involved in the largest number of constraints on other unassigned variables.

Match the explanation of approaches of filtering variable in CSP.

Arc consistency

Propagation from assigned variable to unassigned variable

Both 

Examines further implications, not only 1-step

Forward checking

Detects failures earlier

The time complexity of AC-3 is O(?) because there are ________ edges, take________ time for consistency enforcing, and ________ for arc insertions.

What can be inferred from below?

Match the explanation of well-known problems of planning.

Ramification problem

Representing all things that stay the same

Frame problem

Defining conditions for an action to succeed

Qualification problem

Representing implicit consequences of actions

What is true about linear planning?

It interleaves the goal achievement.

It can lead to Sussman anomaly. 

It is efficient since it does not maintains goal stack.

It is useful when multiple goals interact, since it increases efficiency by separating goals.

Consider PDDL. What is included in what?

Actions

Initial state

Goal

Types

Predicates

Objects

Constants

Domain definition

Problem definition

What is wrong about Bayes Rule?

We can build conditional from its reverse.

Posterior is proportional to likelihood time prior.

Likelihood represents the agent's beliefs given the evidence. 

Posterior and likelihood are a conditional.

Match the type of causal chain with the definition of joint distribution.

Common cause

Common effect

Casual chain