AI

search problem

consists of a state space, a successor function, a start state, and a goal test. A solution is a sequence of actions which transforms the start state to a goal state

goal based agents

from current state to desired future

time complexity

how long does it take to find a solution?

space complexity

how much memory is needed to perform the search?

machine learning

how to acquire a model from data / experience

consistent heuristic

if for every node n and every successor n prime of n generated by any action a, the estimated cost of reaching the goal from n is no greater than the step cost of getting to n prime plus the estimated cost of reaching the goal from n prime.

best first search

instance of tree or graph search which a node is selected for expansion based on an evaluation function. The implementation of best first search is the same as uniform cost search except with heuristic instead of path cost.

preference constraints

instead of absolute constraints, in which a violation rules out a potential solution, many CSPs include preference constraints indicating which solutions are prefered

uniform cost search

instead of expanding the shallowest node like in breadth first search, it expands the node with the lowest path cost. This is done by storing the frontier as a priority queue ordered by cost.

completeness

is the algorithm guaranteed to find a solution when there is one?

when is a constraint graph a tree?

its a tree when any two variables are connected by only one path. Any tree structured CSP can be solved in time linear in the number of variables.

filtering

keep track of domains for unassigned variables and cross off bad options

constraint propagation

making an inference based on the constraints to reduce the number of legal values for a variable

agent function

maps any given percept sequence to an action

branching factor

maximum number of successors of any node

connected components

the components that are connected in a way that makes them dependent on each other.

performance measure

the notion of desirability used to evaluate any given sequence of environment states. It is better to design performance measures according to what one actually wants in the environment, rather than according to how one thinks the agent should behave.

percept

the term to refer to the agents perceptual inputs at any given instant. An agents percept sequence is the complete history of everything the agent has ever perceived.

backtracking search

the term used for a depth first search that chooses values for one variable at a time and backtracks when a variable has no legal values left to assign.

optimality of a star

the tree search version of a star is optimal if the heuristic is admissible, while the graph search version is optimal if the heuristic is consistent. uniform cost search is the same as a star with a null heuristic. The null heuristic is admissible since its 0, so ucs is optimal.

tree decomposition

transforms the CSP into a tree of subproblems and are efficient if the tree width of the constraint graph is small

local consistency

treating each variable as a node in a graph and each binary constraints as an arc, then the process of enforcing local consistency in each part of the graph causes inconsistent values to be eliminated throughout the graph.

greedy best first search

tries to expand the node that is closest to the goal, on the grounds that this is likely to lead to a solution quickly.

disjunctive constraint

two variables must not overlap in time, that one must come first

minimum remaining values heuristic

used for select unassigned variable, chooses the variable with the fewest legal values. Its the most likely to fail first, thereby pruning the search tree

iterative deepening dfs

used in combination with depth first search, finds the best depth limit. It does this by gradually increasing the limit until a goal is found. This will occur when the depth limit is equal to the goal depth.

global constraint

a contraint involving an arbitrary number of variables

partially observable

the sensors on an agent may be noisy and inaccurate, or the parts of the start are missing from the sensor data

select unassigned variable

the simplest strategy is to choose the next unassigned variable in order.

directed arc consistency

A CSP is defined to be directed arc consistent under an order of variable X1, X2,... Xn iff every X is arc consistent with each X for j > i.

forward checking

Cross off values that violate a constraint when added to the existing assignment. forward checking propagates information from assigned to unassigned variables, but doesnt provide early detection for all failures.

three questions

Instead of supplying the CSP with heuristic functions derived from our knowledge of the problem, we can solve the CSP efficiently by answering these three questions. Which variable should be assigned next, and in what order should its values be tried? What inferences should be performed at each step in the search? When the search arrives at an assignment that violates a constraint, can the search avoid repeating this failure?

a star

It evaluates nodes by combining the cost to reach the node with the cost to get from the node to the goal, based on a heuristic.

k consistency

a CSP is k consistent if for any set of k - 1 variables and for any consistent assignment to those variables, a consistent value can always be assigned to any kth variable. 1 consistency says that, give the empty set, we can make any set of one variable consistent

heuristic

a function that estimates how close a state is to a goal

constraint satisfaction problem

a problem in which each state has a set of variable with a value. The problem is solved when each variable has a value that satisfies all the constraints on the variable.

commutativity

a problem is commutative if the order of application of any given set of actions has no effect on the outcome. CSP's are commutative because when assigning values to variables, we reach the same partial assignment regardless of the order.

autonomy

a rational agent should be autonomous. It should learn what it can to compensate for partial or incorrect prior knowledge. After sufficient experience of its environment, the behavior of a rational agent can become effectively independent of its prior knowledge.

forward checking

a simple form of inference. Whenever a variable is assigned, the forward checking process establishes arc consistency for it. For each unassigned variable that is connected to the assigned var by a constraint, delete from the variables domain any value that is inconsistent with the value chosen for the variable.

breadth first search

a simple strategy in which the root node is expanded first, then all the successors of the root node are expanded next, then their successors,

node consistency

a single node is node consistent if all the values in the variables domain satisfy the variables unary constraints

arc consistency

a variable CSP is arc consistent if every value in its domain satisfies the variables binary constraints.

maintaining arc consistency

algorithm that detects inconsistency. After a variable x is assigned a variable, the INFERENCE procedure calls AC3, but instead of a queue of all arcs in the CSP, we start with only the arcs for all y that are unassigned variables neighboring x. From there AC3 does constraint propogation in the usual way, and if any variable has its domain reduced to the empty set, the call to AC3 fails and we know to backtrack immediately.

uninformed search

algorithms that are given no information about the problem other than its definition.

depth first search

always expands the deepest node in the current frontier.

fully observable

an agents sensors give it access to the complete state of the environment at each point in time. A task environment is effectively fully observable if the sensors detect all aspects that are relevant to the choice of the action

partial assignment

an assignment that assigns values to only some of the variables

consistent

an assignment that does not violate any constraint

omniscience

an omniscient agent knows the actual outcome of its actions and can act accordingly, but omniscience is impossible in reality. This shows rationality maximizes expected performance.

agent

anything that can be viewed as perceiving its environment through sensors and acting upon that environment through actuators.

naive bayes

assume all features are independent effects of the label

model based classification

build a model where both the output label and input features are random variables

cutset conditioning

can reduce a general CSP to a tree structured one and is quite efficient if a small cutset can be found

maximum likelihood principle

choose the parameters that maximize the probability of the observed data

optimality

does the strategy find the optimal solution?

information gathering

doing actions in order to modify future percepts. An important part of rationality. Provided by exploration that must be undertaken.

assignment

each state in a CSP is defined by values to some or all of the variables

greedy search

expand a node that you think is closest to a goal state. The problem is it doesn't take the real path cost into account, it takes the heuristic(expected) cost into account.

overfitting

fitting the training data very closely, but not generalizing well

admissible heuristic

never overestimates the cost to reach the goal. In a star, because f = g + h, this is the actual cost to reach the node plus a cost at most equal to the real cost to reach the goal, so it is at most the smallest value to reach the goal.

depth limited search

nodes at a specified depth are treated as if they have no successors, which helps the problem that depth first search has in infinite depth space. This is a problem if we choose a depth which is shallower than the goal depth.

constraint graph

nodes of the graph correspond to variables of the problem. nodes are variables, arcs show constraints.

least constraining value heuristic

once a variable has been selected, the algorithm must decide on the order in which to examine its values. It prefers the value that rules out the fewest choices for the neighboring variables in the constraint graph.

rational agent

one that does the right thing, conceptually speaking, every entry in the table for the agent function is filled out correctly. For each possible percept sequence, a rational agent should select an action that is expected to maximize its performance measure, given the evidence provided by the percept sequence and whatever built in knowledge the agent has.

informed search

one that uses problem specific knowledge beyond the definition of the problem itself. It can find solutions more efficiently than can an uninformed strategy.

path consistency

path consistency tightens the binary constraints by using implicit constraints that are inferred by looking at triples of variables.

PEAS

performance measure, environment, actuators, sensors

binary constraint

related two variables

unary constraint

restricts the value of a single variable

reflex agents

select actions on the basis of the current percept, ignoring the rest of the percept history

depth

shallowest goal node from the root

degree heuristic

used in select unassigned variable. Attempts to reduce the branching factor on future choices by selecting the variable that is involved in the largest number of constraints on other unassigned variables.

precedence constraints

whenever one task must occur before another task

task environments

which are essentially the problems to which rational agents are the solutions