CMSC 471
Depth Limited Search
A depth first search with a predetermined depth limit. Incomplete if limit is less than depth. Not optimal if limit is greater than depth.
Iterative Deepening Search
A depth limited search that gradually increases the limit. Generates states multiple times.
admissible heuristic
A heuristic that never overestimates the cost to reach the goal.
Consistent Heuristic
A heuristic ℎh is consistent if its value is nondecreasing along a path. Mathematically, a heuristic ℎh is consistent if for every node 𝑛n of a parent node 𝑝p,
Alpha-Beta Pruning
A method for eliminating branches during the search that will not affect the outcome; Minimax with Alpha-Beta pruning always returns the same scores as Minimax would when using the same depth limit; Potentially much faster than minimax
Best-first search
A node is selected for expansion based on an evaluation function 𝑓(𝑛)f(n)
Thinking Rationally
AI is not here yet
Uninformed Search
Algorithms that are given no information about the problem other than its definition.
What are characteristics of Alpha Beta Pruning
Alpha-beta pruning uses recursion and DFS strategy to send back up utility values from the terminal (leaf) nodes Alpha-beta pruning algorithm keeps track of alpha and beta values for the nodes that it visits and uses them to prune the search tree. Alpha is updated at Max nodes and beta is updated at Min nodes
What is one state of representation?
Atomic
Optimality of A* search
A∗ is complete . The tree-search version of A* is optimal if ℎ(𝑛)h(n) is admissible, while the graph-version is optimal if ℎ(𝑛)h(n) is consistent A∗ has a high space complexity. It runs out of memory pretty quickly.
What is the best example of the Thinking Humanly approach in AI?
Cognitive Modeling
Thinking Humanly
Cognitive Science
BFS performance
Complete; optimal if step costs are equal; Not space efficient because you visit every node at every level; O(b^d) Time complexity is O(b^d)
Depth Limited Search performance
Completeness? No, if 𝑙<𝑑l<d 𝑙l: limit, 𝑑d: the depth of the shallowest solution. Optimality? No, if 𝑙>𝑑l>d 𝑙l: limit, 𝑑d: the depth of the shallowest solution.
Iterative Deepening Search Performance
Completeness? Yes, if b is finite. Optimality? Yes, if the path cost is a nondecreasing function of the depth of the node. Time Complexity? 𝑂(𝑏𝑑)O(bd) b: branching factor, d: the depth of the shallowest solution. Space Complexity? 𝑂(𝑏𝑑)O(bd) b: branching factor, d: the depth of the shallowest solution.
The minimax algorithm uses a strategy for a search similar to
DFS
What is the search algorithm that is used in solving constraint search problems?
DFS
optimality
Does the strategy find the optimal solution? Lowest path cost among all solutions.
A* search
Expand the node that has the minimum value of 𝑓(𝑛)f(n) 𝑓(𝑛)=𝑔(𝑛)+ℎ(𝑛)f(n)=g(n)+h(n) 𝑔(𝑛)g(n) the cost from the start state to the current node ℎ(𝑛)h(n) the estimated cost from the current node to the goal
greedy best-first search
Expands the node that is closest to the goal. Incomplete even if finite f(n) = h(n)
What type of state representation is used in CSP?
Factored
"Least Constraining Value" chooses the variable with minimum number of constraints. T/F
False
A* search and Greedy best-first search both use the same evaluation function f(n) T/F
False
Arc COnsistency keeps track of remaining legal values for the unassigned T/F
False
Backtracking guarantees to find a solution for CSP. T/F
False
C_xy is arc-consistent with respect to variable x if for every value in the domain D_x all values in the domain D_y satisfy the binary constraint on the arc (x, y). T/F
False
Gradient descent is an optimization algorithm but is never used in Machine Learning due to its problems. T/F
False
Greedy Best First Search guarantees both completeness and optimality T/F
False
In Alpha-beta pruning algorithm, you initialize Alpha with +infinity and Beta with -infinity T/F
False
Local maximum is a point where no other point in the entire search space may have a higher objective function than that. T/F
False
Minimum Remaining Values (MRV)" chooses the remaining values that are minimum in the variable domain. T/F
False
The space complexity of DFS is O(b^m) - b to the power of m - where b is branching factor and m is the maximum depth of the search tree T/F
False
Unary constraint restricts the value of two variables. T/F
False
You can use adversarial search in a single-player puzzle game such as Sudoku, or Rubik's cube T/F
False
How do you express this statement in First-Order Logic? Choose the best answer only. No person buys an expensive policy.
For all X For all y(Person(x) and Policy(y) and Expensive(y)) implies not Buys(x,y)
How do you express this statement in First Order Logic? All birds except penguin fly.
For all birds x (bird(x) and -penguin(x)) implies fly(x)
How do you express this statement in First Order Logic? Select all that apply. No robot is human.
For all x Robot(x) implies not Human(X) for all x HUman(x) implies not Robot(x) There does not exists x Robot(x) and Human(X)
What are properties that apply to the chess game environment?
Fully observable Deterministic Semi Dynamic (when played with a clock) Discrete
Task Environment Dimensions and Properties
Fully or partially observable Single-agent or multi-agent Deterministic or stochastic Static or dynamic Discrete or continuous Known or unknown Episodic or Sequential
What is NOT a task environment property
Goal-based vs utility based
time complexity
How long does it take to find a solution?
Space Complexity
How much memory is needed to perform the search?
Which of the following statements are true in the context of state representation? Select all that apply
In an atomic representation, each state of the world is indivisible - it has no internal structure. In a factored representation, each state of the world can be split to a fixed set of variables or attributes. Search algorithms like BFS and DFS work with atomic representations.
Constraint Satisfaction Problem
In which we see how treating states as more than just little black boxes leads to the invention of a range of powerful new search methods and a deeper understanding of problem structure and complexity.
Informed Search
Informed methods are given additional information about the goal; Greedy, A* search
Acting Rationally
Intelligent Agents
completeness
Is the algorithm guaranteed to find a solution when there is one?
What does space complexity deal with?
Memory only
DFS Performance
Not complete for tree version Not optimal Time complexity is O(b^m) Space complexity is O(b*m)
Search environment properties
Observable Discrete Known Deterministic
Search agents
One specific kind of agents that uses with atomic representations on discrete environments.
Contraposition example
P implies not Q is equal to Q implies not P
Assumptions of Adversarial Search
Perfect information, i.e. both players have access to complete information about the states. Zero-sum , i.e. the utility values at the end of the game are equal and opposite. The search environment is fully observable and deterministic.
Task Environment Specification PEAS
Performance Measure External Environment Actuators Sensors
Unit Resolution Example
Premise:A or B, not B Conclusion: A
P or F Unsatisfiable, satisfiable or Valid?
Satisfiable
Student did NOT pass the test. /// Passed (Test) is a boolean statement. Unsatisfiable, satisfiable or Valid?
Satisfiable
Student passed the test. /// Passed (Test) is a boolean statement. Unsatisfiable, satisfiable or Valid?
Satisfiable
T and (T and A) Unsatisfiable, satisfiable or Valid?
Satisfiable
Adversarial Search
Searching in the presence of an adversary adds uncertainty, because you don't know what the opponent will do
Types of Agents
Simple reflex agents Model-based reflex agents Goal-based agents Utility-based agents
intelligent agent
Special-purpose knowledge-based information system that accomplishes specific tasks on behalf of its users
What does rationality depend on
The performance measure that defines the criterion of success. The agent's prior knowledge of the environment The actions that the agent can perform The agent's percept sequence to date
How do you express this statement in First Order Logic? Choose the best answer only. Some kids like candy.
There exists x Kid(X) and Likes(x,candy)
A partial assignment is one that assigns values to only some of the variables. T/F
True
A solution must be a complete and consistent assignment. T/F
True
A variable in a CSP is arc-consistent if every value in its domain satisfies the variable's binary constraints. T/F
True
A* usually runs out of space long before it runs out of time. For this reason, A* is not practical for many large-scale problems. T/F
True
An admissible heuristic always underestimates the true cost of getting from a state to the goal state. T/F
True
An assignment that does not violate any constraints is called a consistent or legal assignment T/F
True
An assignment that does not violate any constraints is called a consistent or legal assignment. T/F
True
BFS is complete if the branching factor b is finite T/F
True
Backtracking algorithm is a CSP solver T/F
True
Backtracking has inefficiency issue because it may explore areas of the search space that aren't likely to succeed. T/F
True
Both BFS and DFS are both uninformed search T/F
True
Both versions of DFS (tree-search and graph-search) are non-optimal. T/F
True
C_xy is arc-consistent with respect to variable x if for every value in the domain D_x there is some value in the domain D_y that satisfies the binary constraint on the arc (x, y). T/F
True
Comparing AI, Machine Learning, and Deep Learning, one can argue that there is a superset-subset relationship between them such that Deep Learning is a subset of Machine Learning approaches, and Machine Learning is a subset of the broad field of approaches, algorithms and techniques in AI. T/F
True
Consistency is a stronger condition than admissibility T/F
True
Depth-first tree search needs to store only a single path from the root to a leaf node, along with the remaining unexpected sibling nodes for each node on the path, and that is why DFS has a lower space complexity than BFS. T/F
True
Every consistent heuristic is also admissible. T/F
True
Hill climbing may get stuck in local maximum. T/F
True
Hill climbing may get stuck in plateau - flat area T/F
True
IDA*, Recursive Best First Search (RBFS) and SMA* are memory-bounded heuristic searches that try to overcome the issue with A* space complexity - high memory usage. T/F
True
In Genetic Algorithm, each location in the offspring (after crossover) is subject to random mutation with a small independent probability. T/F
True
Intelligence is concerned with mainly rational action T/F
True
Iterative deepening DFS combines the benefits of DFS and BFS. Its space complexity is O(bm) like DFS and it is complete like BFS if the branching factor is finite. T/F
True
Minimum Remaining Values (MRV), Most Constraining Variable and Least Constraining Value are among the heuristics used to make backtracking more efficient. T/F
True
Most Constraining Variable" chooses the variable involved in largest number of constraints on remaining variables. T/F
True
Random restart is one way to overcome hill climbing failure. T/F
True
Rationality is NOT omniscience. Omniscience is impossible in reality T/F
True
Selection, crossover and mutation are all random processes T/F
True
Simulated annealing sometimes allows for bad moves based on a probability that decreases over time T/F
True
Space complexity of DFS is O(bm) where b is the branching factor and m is the maximum depth of the search tree. T/F
True
Space complexity of DFS is lower than BFS. T/F
True
The interdisciplinary field of Cognitive Science brings together computer models from AI and experimental techniques from psychology to construct precise and testable theories of the human mind. T/F
True
The temperature T decreases over time (cools down) to reduce the probability of bad moves by simulated annealing. T/F
True
The term percept refers to the agent's perceptual inputs at any given instant. T/F
True
The time complexity of BFS is O(b^d) where b is branching factor and d is the depth of the shallowest solution. T/F
True
The tree-search version of A* is optimal if h(n) is admissible, while the graph-search version is optimal if h(n) is consistent. T/F
True
The tree-search version of DFS is not complete even if the branching factor b and the depth d are finite. T/F
True
Using heuristics may detect failure of a path earlier. T/F
True
When a CSP is not arc consistent, we may make it arc consistent by using the AC3 algorithm with no guarantee for all problems. T/F
True
there does not exist and x P(x) = for all X -P(x)
True
Acting Humanly
Turing Test
What is the best example of "Acting Humanly"?
Turing Test
Which of the following is NOT an assumption about the search environment?
Unknown
p and not p Unsatisfiable, satisfiable or Valid?
Unsatisfiable
Informed Search
Uses problem-specific knowledge beyond the definition of the problem itself.
-P or T Unsatisfiable, satisfiable or Valid?
Valid
P--> P Unsatisfiable, satisfiable or Valid?
Valid
Breadth First Search
Visits the neighbor vertices before visiting the child vertices Often used to find the shortest path from one vertex to another. A queue is usually implemented
(P or Q) And R CNF?
Yes
P And Q CNF?
Yes
P or Q CNF?
Yes
percept
a mental concept that is developed as a consequence of the process of perception
Depth First Search
a search in which children of a node are considered (recursively) before siblings are considered.
zero-sum assumption
equal and opposite payoff for each player
heuristic function
estimated cost of the cheapest path from the state at node n to a goal state - this is non-negative and problem specific.
What is the A* evaluation function?
f(n) = g(n) + h(n)
Factored state representation
internal structure, although exactly what will depend on the problem
minimax tree
levels switch from max and min nodes
If h1 and h2 are both admissible heuristics, which of the following heuristics are guaranteed to be admissible.
max(h1, h2) min(h1, h2)
Atomic State representation
qualitative measure of how much "internal structure" those models have, from least to most.
structured state representation
relations either of components of the model to itself, or components of the model to components of the environment.
alpha entails beta if and only if alpha implies beta is valid T/F
true
for all x P(x) - there does not exist x not p(x)
true
not for all X P(x) = there exists x not P(X)
true
there exists P(x) = not for all x not p(x)
true