CSCE 420 Final
individuation
division into distinct objects
local consistency
if we treat each variable as a node in a graph and each binary constraint 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.
vanishing point
family of straight lines with directino
cognitive science
field of study that examines how humans and other animals acquire, process, store, and retrieve information, stemmed from the development of computer modeling
three types of queues
fifo, lifo, and priority. fifo is first in first out, lifo is last in first out (aka stack), priority pops off highest priority
toy problem
illustrates or exercises various problem-solving methods.
logical omniscience
if an agent knows a set of axioms, then it nows all consequences of those axioms.
closed-world assumption
if it's not explicitly true, assume false
open-world assumption
if not explicitly true and not explicitly false, it's unknown
static vs dynamic
if the environment can change while the agent is deliberating, the environment is dynamic. Otherwise, static
unknown
if the environment is unknown is has no choice but to take a random action
false negative
if the hypothesis says it should be negative but in fact it is positive.
false positive
if the hypothesis says it should be positive but in fact is negative
cooperative
opposes competitive multi-agent environments, ie avoiding collisions helps the performance measure of all people on the road
Constraint optimization problem
optimally solving CSP's with preferences rather than only requirements.
vlsi layout
position millions of components to minimize area, circuit delay, etc.
quiescent
positions are are unlikely to exhibit wild swings in value in the near future
What are the two choices for representing categories in first-order logic:
predicates and objects
least-constraining value (CSP)
prefers the value that rules out the fewest choices for the neignboring variables in the constraint graph.
objectivist view
probabilities are real aspects of the universe-propensities of objects to behave in certain ways-
n-gram model
probability distribution of n-letter sequences
decision theory =
probability theory + utility theory
relaxed problem
problem with fewer restrictions on the actions
Planning systems
problem-solving algorithms that operate on explicit propositional or relational representations of states and actions
regularization
process of explicitly penalizing complex hypothises
search
process of looking for a sequence of actions
decision tree
represents a function that takes as input a vector of attribute values and returns a 'decision"- a single output value
continuous-valued output attributes
regression tree rather than a classification tree, each leaf is a linear function of some subset of numerical attributes.
behaviorism
rejects any theory involving mental process on the grounds that introspection could not provide reliable evidence
binary constraint
relates two variables in a CSP
factoring
removal of multiple copies of literals
pruned
removed because unwated
belief state
representation of the set of all posible world state it might be in
occlusion
some parts are hidden from some viewing direction
utility
some ways of reaching the goal are better than others. goal is binary, either "unhappy" or "happy". Utility is a spectrum of happiness levels.
singularity
something peculiar or unique, used in the book to refer to futurist's believe in a singularity in which humans perform superhumanly
agent
something that acts
intuition pump
something that amplifies one's prior intuitions
epiphenomenal
something that happens, but casts no shadow, as it were, on the observable world
randomized behavior
sometimes rational in competitive environments because it avoids predictability
weight space
space defined by all possible settings of the weights
holdout cross-validation
splits data, trying to gain most information
backoff model
start by estimating n-gram counts, but for any particular sequence that has a low count, back off to an n-1 gram
Monte Carlo Simulation
start with a search algorithm, from a start position, have the algorithm play thousands of games against itself, using random dice rolls. gives a good approximation of the value of a position.
complete-state formulation
starts with 8 queens and moves them around.
initial state
state the agent starts in
nodes
states in the state space of a problem, usually used in search trees
intentional states
states such as believing, knowing, desiring that refer to some aspect of the external world
probabilistic inference
The computation of posterior probabilities for query propositions given observed evidence
explored set
The set of all states that have been expanded. AKA closed list.
The definition the book chooses to define the goal of AI
act rationally
logicist
hopes to build on logical systems to create intelligent systems
performance measure
"If the sequence is desirable, then the agent has performed well." This is measured by a performance measure
empiricism
"Nothing is in the understanding which was not first in the senses." Locke
physical symbol system hypothesis
"a physical symbol system has the necessary and sufficient means for general intelligent action."
training set of N example input-output pairs
(x1,y1,)(x2,y2),... where each y is generated from unknown function y = f(x), discover approximation function
Turing Test
A computer passes if the human cannot tell whether the written responses come from a person or computer. designed to provide a satisfactory operational definition of intelligence
transition model
A description of what each action does. return the state that results from doing action a in s.
definite clause
A disjunction of literals of which exactly one is positive
probability density function
A function used to compute probabilities for a continuous random variable. The area under the graph of a probability density function over an interval represents probability.
admissible heuristic
A heuristic that never overestimates the cost to reach the goal.
overfitting
A hypothesis is said to be overfit if its prediction performance on the training data is overoptimistic compared to that on unseen data. It presents itself in complicated decision boundaries that depend strongly on individual training examples.
back-propagation
A process by which learning can occur in a connectionist network, in which an error signal is transmitted backward through the network. This backward-transmitted error signal provides the information needed to adjust the weights in the network to achieve the correct output signal for a stimulus.
Satisfiability
A sentence is satisfiable if it is true in, or satisfied by, some model.
perceptron
A single, simple, artificial neuron.
repeated state
A state that is exactly the same as a state previously expanded. Lead to by a loopy path.
human-level AI
AI should strive for "machines that think, that learn and the create."
Game playing
AI's can do that.
autonomous planning and scheduling
AI's can do that.
assumption-based truth maintenance system
ATMS, as opposed to JTMS, represents all the states that have ever been considered at the same time.
limited rationality
Acting appropriately when there is not enough time to do all the computations one might like.
percept
Agent's perceptual inputs at any given instant
fully observable vs partially observable
An agent's sensors give it access to the complete state of the environment at each point in time.
Friendly AI
An artificial intelligence (AI) that has a positive rather than negative effect on humanity.
friendly ai
An artificial intelligence (AI) that has a positive rather than negative effect on humanity.
chronological backtracking
most recent decision point is revisited
DPLL
At each iteration, either eliminate pure literals, unit clauses, or do the split function. For the split function, choose a variable and delete clauses with that variable and delete its negation from clauses. If there is an empty set, return false. If there is an empty clause (you get down to nothing) then it returns true.
Where does reasoning take place?
At the level of categories.
goal formulation
Based on the current situation and the agent's performance measure. The first step of problem solving.
comfirmation theory
Carnap and Hempel, attempted to analyze the acquisition of knowledge from experiences
connectionist models
Cognitive processes depend on patterns of activation in highly interconnected computational networks that resemble neural networks.
expert systems
Computerized advisory programs that imitate the reasoning processes of experts in solving difficult problems
Random-restart hill climbing
Conducts a series of hill-climbing searches from randomly generated initial states, until a goal is found.
global constraint
Constraint involving an arbitrary number of variables
dualist
Descartes considered the mind's activity of thinking and the physical processes of the body to exist in separate realms, what we would now call the dualist theory.
optimality
Does the strategy find the optimal solution?
information gathering
Doing actions in order to modify future percepts
heuristic function
Estimated cost of the cheapest path form the state as node to a goal state.
game theory
Evaluates alternate strategies when outcome depends not only on each individual's strategy but also that of others, does not offer an unambiguous prescription for selecting actions unlike decision theory
best-first search
Expands nodes base on evaluation function. Use priority queue with estimated cost.
adalines
Hebb's learning methods were enhanced by Widrow, and this is what Widrow called his networks.
joint probability distribution
Gives probability of all combinations of values of all variables
Shannon entropy of loaded coin, 99% heads.
H(Loaded) = - (0.99log0.99 + 0.01log.01) = .08 bits
time complexity
How long does it take to find a solution?
space complexity
How much memory is needed to perform the search?
Intelligence
How we think
modus ponens
If P then Q P Therefore Q
subcategory
If basketballs is contained in balls, basketballs is a subcategory of balls
deterministics vs stochastic
If the next state of the environment is completely determined by the current state and the action executed by the agent, the environment is deterministic; otherwise, stochastic. most real-world scenarios should be treated as stochastic.
inheritance
In the object-oriented data model, the ability of an object to inherit the data structure and methods of the classes above it in the class hierarchy. See also class hierarchy.
Completeness
Is the algorithm guaranteed to find a solution when there is one?
In propositional logic, what does the model do?
It fixes the truth value - true or false - for every propositional symbol
model-based, utility-based agent
Looks at sensor and internal state, looks at what will happen if I do this action, judges its utility and uses that to decide on an action.
agent function
Maps any given percept sequence to an action
branching factor
Maximum number of successors of any node
Is there an algorithm that can solve general nonlinear constraints on integer variables?
NO!
constraint graph
Nodes correspond to variables, links connect variables that participate in a constraint
sensorless (conformant) problem
Non-observable Agent may have no idea where it is; solution is a sequence
independence
P(A|B) = P(A) or P(A^B) = P(a)P(b)
bayes rule
P(A|B) = P(B|A)P(A)/P(B)
product rule
P(A ^ B) = P(A|B)(P(B))
causal direction
P(Cause|effect)
Diagnosal direction
P(Effect|Cause)
conditioning
P(Y) = sum of (P(Y|z))P(Z)
inclusion-exclusion principle
P(a v b) = P(a) + P(b) - P(a ^ b)
conditional independence
P(tootache ^ catch|cavity) = P(Tootache|cavity)P(Catch|Cavity)
Four letter acronym composing the task environment
PEAS - performance measure, environment, actuators, sensors
unary constraint
Restricts the value of a single variable
alpha-beta pruning
Returns same move as MiniMax, but prunes branches that won't lead to final decision.
forward pruning
Some moves at a given node are pruned immediately without further consideration
knowledge aquisition
Storage of information in long-term memory
Deduction Theorem
The Deduction Theorem states that a sentence φ logically entails a sentence ψ if and only if the sentence (φ ⇒ ψ) is valid.
speech recognition
The process by which computers recognize voice patterns and words, and then convert them to digital data.
simple reflex agent
These agents select actions on the basis of the current percept, ignoring the rest of the percept history
Definitions of Aritificial Intelligence in four categories
Thinking humanly, thinking rationally, acting humanly, acting rationally
True or False: Every sentence of propositional logic is logically equivalent to a conjunction of clauses
True
What are the two propositional symbols that never change?
True and False
minimax decision
action a1 is the optimal choice fro max because it leads to the state with the highest minimax value.
Is commutativity a property of all CSP's?
Yes
Is forward chaining both sound and complete?
Yes
Can a discrete domain be infinite?
Yes, ie the set of integers or string
MAC (Maintaining ARC Consistency)
a CSP algorithm that detects inconsistent arcs.
n-gram model
a Markov chain of order n - 1. Recall that in a Markov chain the probability of character ci depends only on the immediately preceding characters, not on any other characters. In other words, ni_ is a 3-letter n-gram.
ambiguous
a characteristic of all natural languages
naive bayes model
a commonly occuring patterin in which a signel cause directly influences a number of effects, all of which are conditionally independent, given the cause.
k-consistency
a csp is k-consistent if for any set of k-1 variable and for any consistent assignment to those variables, a co nsistent value can always be assigned to the kth
Horn clause
a disjunction of predicates in which at most one of the predicates is positive, not negated
Markov decision process
a formal class of sequential decision problems
fluent
a function or relation that can vary from one situation to the next. ie At(x,l,s).
consistency
a heuristic is consistent if, for every node n and every successor n' of n generated by an action a, the estimated cost of reaching the goal from n is no greater than the step cost of getting to n' plus the estimated cost of reaching the goal from n'.
generalizes
a hypothesis generalizes well if it correctly predicts the value of y for novel examples.
epistemological commitments
a logical agent believes each sentence to be true or false or has no opinion
hill-climbing search algorithm
a loop that continually moves in the direction of increasing value, uphill.
objective function
a mathematical expression that describes the problem's objective
functionalism
a mental state is any intermediate causal condition between input and output
circumpscription
a more powerful and precise version of the closed-world assumption, the idea is to specify particular predicates that are assumed to as "false as possible"- that is, false for every object except for those for which they are known to be true.
leaf node
a node with no children
authority
a page in the set is considered an authority to the degree that other pages in the relevant set point to it.
hub
a page is considered a hub to the degree that it points to other authoritative pages in the relevant set.
foreshortening
a pattern viewed at a slant can be distorted
factored representation
a possible world is represented by a set of variable/variable pairs
sensor model
a probability distribution over the evidence that its sensor provide, given a state of the world
sensor model
a probability distribution over the evidence that its sensors provide, given a state of the world.
full joint probability distributino
a probability model is completely determined by the joint distribution for all of the random variables
contraint satisfaction problem
a problem is solved when each variable of a factored representation has a value that satisfies all the constraints on the variable.
Hebbian learning
a rule invented by Donald Hebb that modified connection strength between neurons
validity (tautology)
a sentence is valid if it is true in all models
Leave-one-out cross-validation
a set of m training instances is repeatedly divided into an m-1 training set and a test set of 1 instance
model
a set of objects and interpretation and an interpretation that maps each name to the appropriate object, relation, or function
explanations
a set of sentences e such that e entails p.
factored representation
a set of variables, each of which has a value
resolution
a single inference rule that yields a complete inference algorithm when couples with any complete search algorithm b
unit clause
a single literal, a disjunction of one literal
constraint propagation
a specific type of inference that uses constraints to reduce the number of legal values for a variable, which can do the same for another variable, and so on.
factored representation of agent program
a state consists of a vector of attribute values
structured representation
a state includes objects, each of which may have attributes of its own as well as relationships to other objects
atomic representation of agent program
a state is a black box with no internal structure
algorithm
a step-by-step procedure for solving a problem
singular extension
a strategy to mitigate the horizon effect, a move that is clearly better than all other moves in a given position. remembered when discovered at any point in the search of a tree.
pure symbol
a symbol that always appears with the same sign in all clauses
rationality
a system is rational if it does the right thing
bootstrap
a technique of loading a program into a computer by means of a few initial instructions which enable the introduction of the rest of the program from an input device.
procedural attachment
a technique whereby a query about a certain relation results in a call to a special procedure designed for that relation rather than a general inference algorithm
CSP arc-consistent
a variable is AC if every value in its domain satisfies the variable's binary constraints
multiagent planning
achieve your goal with help or hindrance of others
serendipity
accidental success
three choices the online agent has on how closely to monitor the environment
action monitoring (before executing, verify all preconditions hold), plan monitoring (before executing an action, verfy that the remaining plan will still succeed), goal monitoring(before executing an action, check to see if there is a better set of goals it oculd be trying to achieve.)
HLA
action that contain other actions, ie "go to san francisco airport"
angelic nondeterminism
agent makes the choices
software agents (software robots/softbots)
agents in software, on the internet usually
Learning agents
agents that are built and then improve because they can learn
machine translation
ai can do that
logistics planning
ai's can do that
robotics
ai's can do that
spam fighting
ai's can do that
pure optimization problems
aim is to find the best state according to an objective function
Manhattan distance
aka city block distance, the sum of the horizontal and vertical distances.
uninformed search algorithm
algorithms that are given no information about the problem other than its definition.
constructive induction algorithms
algorithms that can generate new predicates
logical positivism
all knowledge can be characterized by logical theories connected to observations sentence, combines rationality and empiricism
predecessors of s ate x
all those states that have x as a successor
contingent plans
allow the agent to sense the world during execution to decide what branch of the plan to follow
machine learning
allows computer to adapt to new circumstances and to detect and extrapolate patterns
knowledge representation
allows computer to store what it knows or hears
angelic semantics
allows provably correct high-level plans to be derived without consideration of lower-level implementations
contingent plans
allows the agent to sense the world during execution to decide what branch of the plan to follow
Hierarchical task network (HTN)
allows the agent to take advice from the domain designer in the form of HLA, that can be implemented in various ways by lower-level sequences.
Hierarchical task network
allows the agent to take advice from the domain designer in the form of high-level actions that can be implemented in various ways by lower-level action sequences.
alpha-beta pruning
alpha initialized to negative infinity at a node, beta to positive infinity.
optimal
always finds a global minimum/maximum
loss function
amount of utility lost predicting h(x) = y when the correct answer is f(x) = y.
demonic nondeterminism
an adversary makes the choices
reconstruction
an agent builds a geometric model of the world from an image or a set of images.
recognition approach
an agent draws distinctions among the objects it encounters based on visual nad other information
fundamental idea of decision theory:
an agent is rational iff it chooses the action that yields the highest expected utility, averaged over all the possible outcomes of the action. (MEU, maximum expected utility)
model-based agent
an agent that uses a model
uncertain
an environment is uncertain if it is not fully observable or not deterministic
frames
an idea by Minsky (1975), people followed up with a more structured approach of assembling facts about particular object and event types and arranging the types in a large taxonomic hierarchy, like a biological taxonomy
completeness
an inference algorithm is complete if it can derive any sentence that is entailed.
sound/truth-preserving
an inference algorithm that derives only entailed sentences
utility function
an internalization of the performance measure.
computational linguistics/natural language processing
an intersection of modern linguistics and AI
contrained optimization
an optimization problem is constrained if solutions must satisfy some hard constraints on variable values.
resolvent
an ordinary resolution step takes two clauses c1 and c2 and resolves them to produce the resolvent C. An inverse resolution take C and produces C1 and C2.
computational learning theory
analyzes the sample complexity and computational complexity of inductive learning
axiom
another name for sentence, when a sentence is taken as a given without being derived from other sentences
agent
anything that can be viewed as perceiving its environment through sensors and acting upon that environment through actuators
unit clause
appears by itself
discrete vs continuous
applies to the state of the environment, way time is handled, and percepts/actions of the agent. If an environment has a finite number of distinct states, it's discrete. Otherwise continuous.
.theorem proving
applying rules of inference directly to the sentences in our knowledge base to construct a proof of the desired sentence without consulting models.
scaled orthographic projection
approximates distances mathematically
expanding the current state
as in search trees, apply each legal action to the current state, generating new states
spatial substances
as opposed to events, mass nouns
automatic assembly sequencing
assembly of things automatically
path cost
assigns a numerical cost to each path
probability model
associates a numerical probability with each possible world
signifigance test
assume there is no underlying pattern. (null hypothesis). find deviation from random. if greater than 5% deviation from random, probably a significant pattern in data.
expressiveness
atomic, factored, and structured representations lie along this axis. a more expressive representation can capture everything a less expressive one can capture, plus more.
Artificial intelligence
attempts not just to understand but to build intelligent entities
recursive best-first search
attempts to mimic the operation of standard best-first search, but uses a linear space.
line search
attempts to solve the hill problem by finding gradients across a line, often doubling the distance with each step, going until the gradient sign changes, to find either local min or local max (the objective goal).
propositional attitudes
attitudes that an agent can have toward mental objects such as believes, knows, wants, intends, and informs. these attitudes do not behave like normal predicates.
Linear interpolation smoothing
backoff model that combines trigram,bigram, and unigram models by linear interpolation
backjumping CSP
backtracks to the most recent assignment in the conflict set
multiple inheritance
banned in some OOP languages because the inheritance algorithm might find two or more conflicting values answering the query.
event calculus
based on points of time rather than on situations
machine evolution (now called genetic algorithms)
based on the belief that by making an appropriate series of small mutations to a machine-code program, one can generate a program with a good performance for any particular task.
Inverse resolution
based on the observation that if the example classifications follow from background ^ hypothesis ^ descriptions, the one must be able to prove this fact by resolution (because resolution is complete).
uniform-cost search
basically bfs with step-cost, instead of expanding shallowest node expand the node with lowest path cost.
local beam search
begins with k randomly generated states. At each step, all successors are generated, if any one is a goal, it halts. otherwise it repeats.
specular reflectino
behavior of a perfect mirro
superpixels
big pixels, over segmentation o of an image containing hunders of homoegenous regions, these homogenous regions are called superpixels
another name for uninformed search
blind search
corpus
body of text
search tree
branches are actions and nodes correspond to states
cognitive science
brings together computer models from AI and experimental psychology to construct testable theories of the human mind
incremental belief-state search
build up the solution one physical state at a time
certainty factors
calculus of uncertainty developed by MYCIN, a program to diagnose blood infections
planning graph
can be constructed incrementally, starting from the initial state. each layer contains a superset of all the literals or actions that could occur at the time step and encodes mutual exclusion relations amont literals or actions that cannot cooccur.
informed search algorithm
can do well given some guidance on where to look for solutions.
depth-limited search
can handle infinite spaces unlike depth-first, because it has a predetermined depth limit l. nodes at depth l are treated like they have no successors. The problem is that it introduces incompleteness of l < d, the shallowest the goal.
State-space search
can operate in the forward direction (progression) or the backward direction (regression)
decision trees
can represent all boolean functions
computable
capable of being computed
inductive logic
capable of computing the correct probability for any proposition from any collection of observations.
extrinsic
categories not maintained under subdivision, ie. length, height of a person
composite objects
categories of these are often characterized by structural relations among parts
natural kind categories
categories with no clear cut definition
subsumption
checking if one category is a subset of another by comparing their definitions
classification
checking whether an objects belongs in a category
stochastic hill climbing
chooses at random from among the uphill moves
backtracking search with csp
chooses values one variable at a time and backtracks when a variable has no legal values left to assign
model preference logic
circumscription is an example, a sentence is entailed if it is true in all preferred models of the KB, as opposed to the requirement of truth in all models in classical logic.
prioritized circumscription
circumscription with priority
environment class
class of environments. ie used to test driverless taxi, many different traffic and weather varieties, pedestrian concentration, etc.
unit clause
clause which has one literal
goal clauses
clauses with no positive literals
tuple
collection of objects arranged in a fixed order and written with angle brackets surrounding the objects
pattern databases
collection of the exact solution costs for every possible subproblem instance- for the 8 puzzle problem every possible configuration of the four tiles and the blank.
A* search
combines g(n), the cost to reach the node from start, with h(n), the cost to get from the node to to the goal. f(n) = g(n) + h(n),
simulated annealing
combines hill climbing with a random walk in some way that yields both efficiency and completeness. isn't greedy in that it sometimes goes down making it possible to be complete, and doesn't go forever in one direction, which is complete but very ineffient.
Decision theory
combines probability theory with utility theory, provides a formal and complete framework for decisions made under uncertainty
total cost
combines search cost and past cost of the solution found
competitive ratio
comparison of the cost with the path cost of the path the agent would follow if it knew the search space in advance.
loss function
tells us how bad each error is
four ways to measure an algorithm's performance
completeness, optimality, time complexity, space complexity
offline search algorithm
compute a complete solution before setting foot in the real world and then execute the solution.
minimax algorithm
computes the minimax decision from the current state
architecture
computing device with physical sensors and actuators
agent program
concrete implementation running in a physical system, unlike an agent function
conjunctive normal form
conjunction of clauses
beam search
consider only a beam of the n best moves rather than considering all possible moves. this is dangerous because there is no guarantee that the best move will not be pruned away.
utility-based agent
considers not goals (whether or not I've reach the goal determines whether I'm happy or sad), but a spectrum of happiness levels, an internalized performance measure.
Narrow content
considers only the brain state
atomic sentences
consist of a single proposition symbol
constraint hypergraph
consists of ordinary nodes (circles) and hypernodes (squares) which represent n-ary constraints.
evaluation function
construed as a cost estimate, the node with the lowest evaluation is expanded first.
observation sentences
correspond to sensory inputs
disjoint pattern databases
create a lower bound on subproblem complexity by finding the number of possible moves in the subproblem.
model selection
defines the hypothesis space
logical minimization
defining an objects as the smallest one satisfying certain conditions
unconditional or prior probabilities
degrees of belief in propositions in the absence of any other information
internal state
depends on the percept history and thereby reflects at lesat some of the unobserved aspects of the current state
subjectivist/Bayesian
describes proabiltiies as a way of characterizing an agent's beliefs, rather than as having an external physical significance.
goal info
describes situations that are desirable
object model
describes the objects that inhabit the visual world
object modle
describes the objects that inhabit the visual world-people, building, trees, cars, etc.
rendering model
describes the physical, gemoetric, and statistical process that produce the stimulus from the wrodl
rendering model
describes the physical, geometric, and statistical process that produce the stimulus from the world.
truth maintenance system
designed to handle belief revisio
background subtraction
detect background and subtract it
goal test
determines whether a given state is a goal state
homeostatic devices
devices that contain appropriate feedback loops to achieve stable appropriate behavior
nonmonotonic logics
devised with modified notions of truth and entailment in order to capture this behavior
graph
directed network
feature selection
discard attributes that appear to be irrelevant
partition
disjoint exhaustive decompsition
clause
disjunction of literals(thing ored together)
straight-line distance heuristic
draw straight line from x to y, use to inform some search algorithm
justification-based truth maintenance system
each sentence in the knowledge base is annotated with a justification consisting of the set of sentences from which it was inferred
triangle inequality
each side of a triangle cannot be longer than the sum of the two sides
linear constraints
each variable appears in linear form in a constraint equation in CSP
global maximum
elevation corresponds to an objective function in a state-space function, find the global maximum
state-space landscape
elevation corresponds to either cost or the objective function
feature extraction
emphasizes simplme computations applied directly to sensor observations
probability theory
enable a degree of belief
Natural language processing
enables successful communication in english
semidynamic
environment itself doesn't change but performance measure does, it's semidynamic
randomize
escape from infinite loops is possible if the agent can randomize his actions
stochastic games
essentially, a mixture of luck and skill. there is a random element
local search
evaluating and modifying one or more current states rather than systematically exploring paths from an initial state. operate using a single current node and move on to neighbors of that node.
practical ignorance
even if we know everything, we might be uncertain about a particular patient because not every test has been runo
discrete events
events with a definite structure
traveling salesperson problem
everything must be visited exactly once
bunch
example: BunchOf({Apple1,Apple2,Apple3})
redundant paths
exists whenever there is more than one way from one state to another
depth-first search
expands the deepest node in the current frontier of the search tree.
Greedy Best First Search
expands the node that is closest to the goal. on the grounds that this is likely to lead to a solution quickly.
assumptions
explanations can include assumptions-sentences taht are not known to be true, but would suffice to prove p if they were true.
temporal logic
facts hold at particular times and that those times are ordered
constraint learning (CSP)
find a minimum set of variables from the conflict set that causes the problem
route-finding problem
find optimal route
continuous and integer-valued input attributes
find split point with highest information gain
iterative deepening search
finds the best depth limit, general strategy often used with depth first search. combines the benefits of BFS and DFS by iteratively increasingly the depth that it uses for the depth limit. 4. in general, iterative deepening is the preferred uninformed search method when the search space is large and the depth of the solution is not known.
optimization
finds the best hypothesis within a spect
complete local search algorithm
finds the goal if it exists
Existential Instantiation
for any sentece alpha, variable v, and constant k that does not appear elsewhere in the knoweldge base, we can use the existence thing and take out the variable
principle of tichromacy
for any spectral energy density, no matter how complicated, it is possible to construct another spectral energy density consisting of a mixture of just three color-red, gree, nd blue- such that a stupid ****in human can't tell the difference.
definition of a rational agent
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.
default logic
formalism in which defautl rules can be written to generate continuous, nonmonotonic conclusions
Three steps of search
formulate, search, execute
data-driven reasoning
forward chaining is an example of this- reasoning in which the focus of attention starts with the known data
And-elimination
from a conjunction, any of the conjuncts can be inferred (if a and b then a)
lens system
gather sufficient light while keeping the image in focus
upper ontology
general framework of concepts
countours
general outline of cost driven over a search tree, kind of like on a topological map. possible with a star
induction
general rules are acquired by exposure to repeated associations between their elements
weak methods
general-purpose search mechanisms trying to string together elementary reasoning steps to find complete solutions
environment generator
generate environments with certain parameters, likelihood of certain events, etc.
dropping conditions
generates a weaker definition and allows a larger set of positive examples
applicable
given a state s, there is a set of actions. These actions are applicable in s.
text classification/categorization
given a text of some kind, decide which of a predefined set of classes it belongs to.
language identification
given a text, determine what natural language it is written in
protein design
goal is to find a sequence of amino acids that will fold into a 3-d design with the properties to cure a disease
8-queen problems
goal is to place eight queens on a chessboard such that no queen attacks an other.
existential grpahs
graphical notation of nodes along edges
hill climbing is often called a
greedy local search algorithm
population/individual
group of things put there in genetic algorithm. group of tries.
truth maintenance systems
handle knowledge updates and revisions efficiently
conditional/posterior probability
have givens.
early stopping
having an algorithm stop generating nodes when there is no good attribute to split on, rather than going to all the trouble of generating nodes and then pruning them away
semantic
having to do with the meaning of words or language
another named for informed search
heuristic search
Lisp
high-level programming language, the dominant AI programming language
fitness function
higher values for better states
queue
how a frontier needs to be expanded in a way that the search algorithm can easily choose the next node to expand according to its preferred strategy. operations are empty, pop, and insert
learning
how all agents can improve performance
exploration
how an agent discovers an unknown environment
inheritance
how categories serve to organize and simplify the knowledge base
knowledge representation language
how each sentence is expressed, represent some assertion about the world
syntax
how sentences are expressed
taxonomy
how subclass relations organize categories
actuators
how the agent acts on its environment
sensors
how the agent perceives its environment
logical inference
how the definition of entailment can be used to derive conclusions
accessibility relatinos
how the worlds are connected in modal logic, one relation for each modal operator. we say that w1 is accessible from world w0 with respect to ka if everything in w1 is consistent with what A knows in w0.
learning curve/happy graphs
how we can evaluate the accuracy of a learning algorithm, represented by graph in which each point represents a specific number of trials, and on the y axis you have the proportion correeect, as set grows, happiness should increase
decision tree pruning
how we combat overfitting. eliminate nodes that are not relevant
test set
how we measure the accuracy of a hypothesis
consistent hypothesis
hypothesis that agrees with all the data
m is a model of alpha
if a sentence alpha is true in model m
satisfies
if a sentence alpha is true in model m, we say that m satisfies alpha.
ground resolution theorem
if a set of clauses is unsatisfiable, then the resolution closure of those clauses contains the empty clause
first-choice hill climbing
implements stochastic hill climbing by generating successors randomly until one is generated that is better than the current state
agent program
implements the agent function
omniscience
impossible in reality, an omniscient agent knows the actual outcome of its actions and can act accordingly
partial assignment
in CSP solving, it's when some variables are not assigned.
constraint weighting
in CSP's, give weight to constraints. adds more weight to the difficult to solve constraints
precedence constraints
in CSP's, something must happen before something else.
complete assignment
in a constraint satisfaction problem, it's when every variable is assigned a value. A CSP is a consistent, complete assignment.
knowledge-based agents
in ai, an approach to intelligence that states that people learn through processes of reasoning that operate on internal representations of knowledge.
episodic vs sequential
in an episodic task environment, experience is divided into atomic episodes of of the agent receives a percept and performs a single action.
incompleteness theorem
in any formal theory, there are true statements that are undecidable in the sense that they have no proof within the theory.
node-consistent
in csp, a variable is node-consistent if all the values in the domain satisfy its unary constraints.
MRV heuristic
in csp, picks a "most constrained variable" or a "fail first" variable.
skolem constant
in instantiation, what you call something when you don't want it to be mistaken as a variable or something
known vs unkown
in known, outcomes for all actions are given.
competitive
in maximizing your performance measure you seek to minimize that of others
possible worlds
in modal logic, need possible words rather than just one world of truth
alliances
in multiplayer games, these are agreements
reference class problem
in trying to determine the outcome probability of a particular experiment, the frequentist has to place it in a reference class of "similar experiments with known outcome frequencies.
total Turing Test
includes video signal
evidence
information
entropy
information gain , the fundamental quantity in information theory
critic (as related to the learning agent)
informs the learning element on how the agent is doing and determines how the performance element should be modified to do better in the future.
Problem's five components
initial state, possible actions available, description of what each action does, goal test (whether an agent is or isn't in the goal state), path cost(numeric cost to each path)
online search
interleaves computation and action. first it takes an action, then it observes the environment and computes the next action. Good in an environment where there is a penalty for sitting around and computing for too long.
deformation
internal degrees of freedom of othe object changes its appearance.
wide content view
interprets it from the point of view of an omniscient outside observer with access to the whole situation, who can distinguish differences in the world
Bayesian network
invented to allow efficient representation of uncertain knowledge
incremental formulation
involves operators that augment the state description, starting with an empty state.
referential transparency
it doesn't matter what term a logic uses to refer to an object, what matters is the object that the term names.
strongly k-consistent
k-consistent, (k-1) consistent, (k-2) consistent, and so on..
tabu search
keeping a small list of recently visited states and forbidding the algorithm to return to those states
model
knowledge about how the world works
logic
laws of thought supposed to govern the operation of the mind
trying to use logic to cope with a domain like medical diagnosis fails for three main reasons:
laziness, theoretical ignorance, practical ignorance
solution
leads from an action state to a goal state
touring problems
like route finding, but they involve going to a certain number of things, so where you have traversed must be recorded in internal state
robot navigation
like route-finding but continuous
probability
likelihood that a particular event will occur
microworlds
limited problems that appear to require intelligence to solve
mass nouns
linguistic synonym for stuff
count nouns
linguistic synonym for things
probability distribution
list of possible outcomes with associated probabilities
pure literal
literal which appeals with the same truth value throughout a sentence
what features often trap hill-climbing
local maxima, ridges, plateaux
gradient
local slope, often used to find a maximum.
canonical representation
logically equivalent states should map to the same data structure
optimal solution
lowest path cost among all solutions
global minimum
lowest valley in a state-space landscape. if elevation corresponds to cost, this is the goal.
weak AI
machines could act as if they were intelligent
strong AI
machines that do so are actually thinking
current-best-hypothesis
maintain a single hypothesis, and to adjut it as nex examples arrive in order to maintian consistency.
satisficing
making decisions that are "good enough", rather than laboriously calculating an optimal decision
exhaustive decomposition
males and females constitute an exhaustive decomposition of the animals if, if you are not a male you are a female animal.
policy
mapping from every possible state to the best move in that state
model
mathematical abstractions which simply fix the truth or falsehood of every relevant sentence.
utility
mathematical treatment of "preferred outcomes"
diameter of state space
max number at which any node can be reached by any other node
noisy values
may return different values for f(x) each time x occurs
semantics
meaning of sentences
theroetical ifnorance
medical science has no complete theory for a domain
MA*
memory-bounded a star
biological naturalism
mental states are high-level emergent features that are caused by low-level physical processes in the neurons, and it is the properties of the neurons that matter.
large-scale learning
millions of examplees
minimum description length
minimizes the total number of bits required.
hidden Markov Models
models for speech recognition based on a rigorous mathematical theory and generated by a process of training on a large corpus of real speech data
total function
models in first-order logic require total functions, that is, there must be value for every input tuple
language models
models that predict the probability distribution of language expression
chance nodes
must be incorporated along with min and max nodes in stochastic games
multiagent planning
necessary when there are other agents in the environment with which to cooperate or compete
robotics
needed for total turing test,allows to manipulate objects and move about
computer vision
needed for total turing test. allows to perceive objects
constraint language
needed to solve CSP's with infinite domains.
neurons
nerve cells
dead-end
no goal state is achievable
optimally efficient
no other optimal algorithm is guaranteed to expand fewer nodes, A star is an example.
child nodes
nodes that are children of parent nodes (one level below)
process categories/liquid event categories/temporal substances
not discrete events, subcategories of discrete events
tractability
not intractable, ie time required to solive does not grow exponentially with the size of the instances
modal logic
not just concerned with the modality of truth. involves sentences as arguments, rather than terms.
with competitive partially observable games, what is the goal?
not just to move pieces to the right squares but also to minimize the information that the opponent has about their location.
description logics
notations that are designed to make it easier to describe definitions and properties of categories
sample complexity
number of required examples of the hypothesis space
depth
number of steps from root
frequentist
numbers can come only from experiments
probabilistic agent
numerical edegree of belief between 0 and 1
aspect
object look different when seen from different directions
motion blur
objects in the scene that move will apear blurred because they send photons.
communication
often emerges as a rational behavior in multiagent environments
minimum slack algorithm
on each iteration, schedule fo rht earliest possible start whichever unscheduled actions has all its predecessors scheduled and has the least slack
execution
once a solution is found, the actions it recommends can be carried out.
nondeterministic environment
one in which actions are characterized by their possible outcomes, no probabilities are attached to them, usually associated wit performance measures that require the agent to succeed for all possible outcomes of its actions.
complementary literals
one literal is the negation of another
ambiguity
one name for two or mroe different categories
sum of squared differences
one possible measure of similarity
rational agent
one that acts so as to achieve the best outcome, or , given uncertainty, the best expected outcome.
Syllogisms
patterns for argument structures that always yield correct conclusions when given correct premises
real-world problem
people care about it.
what are six risks given by the book that can be caused by AI
people might lose their jobs to automation people might have too much leisure time people might lose their sense of uniqueness AI systems might be used toward undesirable ends the use of AI systems might result in a lack of accountability the success of AI might mean the end of the human race
neats
people who think that AI theories should be grounded in mathematical rigor
diffuse albedo
percent of light you reflect
reward or penalty (learning agent)
performance standard distinguishes part of the incoming percept as reward or penalty, provides direct feedback on the quality of the agent's behavior.
topological sort
pick any variable to be the root of the tree, and choose an ordering of the varibles such that each variable appears after its parent in the tree.
pinhole camera
pinhole opening, o, and an image plane at the back of the box.
specularites
places of perfect or near perfect reflection
wrappers
programs taht extract information from a page
error rate of hypothesis
proportion of mistakes it makes
principle of indifference
propositions that are syntacticalkly "symmetric" with respect to the evidence should be accorded equal probability
description logic
provide a formal language for constructing and combing category definitions and efficient algorithms for deciding subset and superset relationships between categories.
semantic networks
provide graphical aids for visualizing a knowledge base and efficient algorithms for inferring properties of an object on the basis of its category membership
NP-completeness
provides a method to recognize an intractable problem.
closed-world assumption
provides a simple way to avoid having to specify lots of negative information
control theory
purposive behavior arises from a regulatory mechanism trying to minimize error- the difference between a current a goal state.
predicate indexing
putting the things in the right places, eg putting all the Knows in one bucket and Brothers in the other, just so as not to confuse anything.
mutation
random change
depth of filed
range in which you can focus
stochastic beam search
rather than choosing the best k from the pool of candidate successors, it just randomly generates k. Solves the diversity problem
intelligence is mainly concerned with __________
rational action
performance element
responsible for selecting external actions
problem generator (as relates to learning agent)
responsible for suggesting actions that will lead to new and informative experiences. suggests exploratory actions that might be suboptimal in the short run.
learning element
responsible fro making improvements in a learning agent
PAC learning algorithm
returns approximately correct hypothesis
retrograde minimax search
reverse the rules of chess to do unmoves rather than moves. any move by white that, no matter what move black responds with, ends up in a position marked as a win, must also be a win.
inference rules
rules that can be applied to derive a proof
diffuse reflection
scatters light evenly across the direction leaving a surface so the brightness of a diffuse surface doesn't depend on the viewing direction
active sensing
send out a signal, and sense the reflection of this signal off of the environment
path
sequence of state connected by sequence of actions.
decision lists
series of tests, each of which is a conjunction of literals
frontier (aka open list)
set of all leaf nodes available for expansion
hypothesis space
set of all possible hypotheses
sample space
set of all possible words
version space
set of hypothesis remaining
grammar
set of rules of a language
Knowledge base
set of sentences, the central component of a knowledge-based agent
language
set of strings
conflict set (CSP)
set that are in conflict with a variable
Kolmogorov's axioms
showed how to build up the reset of probabiltiy theory from this simple foundation and how to handle the difficulties caused by continuous variables.
simple distinction between four agents.
simple reflex agents repond directly to precepts. model-based reflex agents maintain internal state that aren't evident in current percept. goal-based agents act to achieve goals. utility-based agent try to maximize expected "happiness"
discrete, finite domains
simples CSP's, ie map coloring and scheduling with time limits, eight-queen problem
ockham's razor
simplest hypothesis is preferable
sma*
simplified ma star
underconstrained
solutions are densely distributed
preference constraint (CSP)
solutions are preferred, not necessarily required
safely explorable
some goal state is reachable from every reachable state
instances
strings that match a schema
ontological engineering
study of how to create representations, concentrating on general concepts-such as events,times,physical objects, and beliefs-that occur in many different domains.
operations research
study of how to take rational decisions when payoffs from actions are not immediate but instead result from several actions taken in sequence.
qualitative physics
subfield of AI that investigates how to reason about physical systems without plunging into detailed equations and numerical simulations
schema
substring in which some of the positions can be left unspecified
margnization/summing out
summation of the probabilitiesi for each possible value of the other variables, thereby takeng them out of the equation.
Model-based agent sequence
takes in sensor data, looks at state, analyzes how the world has evolved and what past actions have done, tries to find out more info about unobserved aspects of the state. Makes decision in same way as model-based agent.
intelligent agent
takes the best possible action in a situation
linear regression
task of finding the hw that best fits the data
discretize
technique in local search, solve continuous problems by discretizing neighborhood of each state.
deformable template
tells us which configurations are acceptable: the elbow can bend but the head is never joined to the foot.
ground term
term without variables
PDLL
the Plnning Domain Definition Langugae, describes the initial nad goal states as conjunctions of literals, and actions in terms of the preconditions and effects
game
the actions of one player can significantly affect the utility of another
transhumanism
the active social movement that looks forward to this future in which humans are merged with -or replaced by-robotic and biotech inventions.
unobservable
the agent has no sensors at all, the environment is unobservable
reinforcement learning
the agent learns from a series of reinforcements-rewards or punishments
unsupervised learning
the agent learns patterns in the input even though no explicit feedback is supplied
supervised learning
the agent observes some example input-output pairs and learns a function that maps from input to output.
supervised learning
the available feedback provides the correct answer for example inputs
materialism
the brain's operation according to the laws of physics constitutes the mind
effective branching factor
the branching factor that a uniform tree of depth d would have to have in order to contain n+1 nodes
percept sequence
the complete history of everything the agent has ever perceived
in an inference, what is the right side called (the side pointed to)?
the conclusion or consequent
grounding
the connection between logical reasoning processes and the real environment in which the agent exists.
specializtion
the extension of the hypoteheis must be decreased to exclude the exmaple
realizable
the hypothesis space contains the true function
entailment
the idea that a sentence follows logically from another sentence
belief state
the key concept required for solving partially observable problems. it represents the agent's current blief about the possible physical state it might be in, given the sequence of actions and percepts up to that point.
perceptron convergence theorem
the learning algorithm can adjust the connection strengths of a perceptron to match any input data, provided such a match exists
argument from consciousness
the machine has to be aware of its own mental states and actions (from Can Machines Really think?)
monism/physicalism
the mind is not separate from the body-mental states are physical states
goal predicate
the name of the output for a decision tree
small-scale learning
the number of training examples ranges from dozens to low thousands
critical path
the path whose total duration is longest
What is rational depends on: (four things)
the performance measure (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.
epistemological commitment
the possible stats of knowledge that it allows with respect to each fact
rationalism
the power of reasoning in understanding the world
In an inference, what is the left side called (the side not pointed to)?
the premise or antecedent
Markov chain
the probability of charater ci depends only on the immediately preceding characters, not on any other characters
task environment
the problems to which rational agents are the solutions
information extraction
the process of acquiring knowledge by skimming a text and looking for occurences of a particular class of object and for relatinoships among objects
smoothing
the process of adjusting the frequency of low-frequency counts
segmentation
the process of breaking an image into regions of similar pixels
problem formulation
the process of deciding what actions and state to consider, given a goal
abstraction
the process of removing detail from a representation
horizon effect
the program is facing an opponent's move that causes serious damage and is ultimately unavoidable, but can be temporarily avoided by delaying tactics.
recall
the proportion of all relevant documents in the collection that are in the result set.
Precision
the proportion of documents in the result set that are actually relevant
perplexity
the reciprocal or probability , normalized by sequence length.
state space
the set of all states reachable from the initial state by any sequence of actions.
monotonicity
the set of entailed sentences can only increase as information is added to the knowledge base
boundary set
the set of hypothesis composing the boundary
vocabulary
the set of symbols that make up the corups and the model
iterative-deepening A*
the simplest way to reduce memory requirement for a start is to adapt the idea of iterative deepening to the heuristic search context.
phenomenology
the study of direct experience
neuroscience
the study of the nervous system, particularly the brain.
inverted spectrum thought experiment
the subjective experience of person X when seeing red objects is the same experience that the rest of us experience when seeing green objects, and vice versa.
decentralized planning
the subplan constructed for each body may need to include explicit communicative actions with other bodies
information retrieval
the task of finding documents that are relevant to a user's need for information.
qualification problem in AI
the the inability to capture everything in a set of logical rules
minimax value
the utility of being in the corresponding state, assuming that both players play optimally from there to then end of the game.
expected utility
the utility the agent expects to derive, on average, given the probabilities and utilities of each outcome.
weights
the value of y is changed by changing the relative weight of one term or another
marginal probability
the values in the margins of a joint probability table that provide the probabilities of each event separately
ontological commitments
the world is composed of facts that do or do not hold in any particular case
single vs multiagent
there are multiple agents involved which have a goal of maximizing a performance measure dependent upon other agents
dualism
there is a part of the human mind outside of nature
stationary distribution
there is a probability distribution over examples that remains stationary over time.
intrinsic
things that belong to the very substance of an object
stuff
things that cannot be individuated ie, butter, water, etc.
scruffies
those who would rather try out lots of ideas, write some programs, and assess what seems to be working
path consistency (CsP)
tightens the binary constraints by using implicit constraints that are inferred by looking at triples of variables.
autonomy
to the extent that an agent relies on the prior knowledge of its designer rather than on its own percepts, we say an agent lacks autonomy
laziness
too much work to list complete set of antecedents, etc.
True or false: if kb is true in the real world, then any sentence alpha derived from kb by a sound inference is also true in the real world
true
reification
turning a proposition into an object, ie (Basketball(b)) into basketballs
synonymy
two names for the same category
disjoint
two or more categories are disjoint if they have nothing in common
logical equivalence
two sentences a and b are logically equivalent if they are true in the same set of models.
disjunctive constraint
two things must not overlap in time (CSP)
problem-solving agent
type of goal-based agent that uses atomic representation.
planning agent
type of goal-based agent that uses more advanced factored or structure representations
breadth first search
type of uninformed search, a simple strategy in which the root node is expanded first, then all of the successors are expanded, then their successors, etc.
bidirectional search
type of uninformed search. Two searches: one forward form the root and one backward form the goal. Goal test replaced with frontier intersect test. Requires a method of computing predecessors.
clusternig
type of unsupervised learning: detecting potentially useful clusters of input examples
search cost
typically depends on time complexity but also can include a term for memory usage
missing data (dt)
unrecoded values or data that might be too expensive to obtain
hypothesis
when a learning agent derives an approximation function
automated reasoning
use stored info to answer questions and to draw new conclusions
backed-up value
used by recursive best-first search. as recursion unwinds rbfs replaces f value of each nodel with teh best f value of its children, the backed up value.
condition-acting rule
used by simple reflex agent. If this happens in environment do this.
memoization
used in computer science to speed up progams by saving the results of computation
sensorless/conformant planning
used to construct a plan that works without the need for perception.
goal-directed reasoning
useful for answering questions such as "what shall I do now?" Backward chaining is an example
online planning agent
uses execution monitoring and splices in repairs as needed to recover from unexpected situations,
online planning agent
uses execution monitoring and splices in repairs as needed to recover from unexpected situations.
goal-based agent
uses goal information
goal-based, model-based agent
uses internal state and current sensor readings to look at what will happen if I do x, then looks at the series of goals to define the action.
dominated
when a task is the same as another task B, but B is cheaper and faster.
backtracking search
variant of depth-first search, uses less memory. only one successor is generated at a time, each partially expanded node remembers which successor to generate next.
genetic algorithm
variant of stochastic beam search in which successor states are generated by combining two parents states rather than by modifying a single state.
robotic vehicles
vehicles that are robotic
cognitive psychology
views the brain as an information-processing device
higher-order logic
views the relations and functions referred to by first-order logic as objects in themselves
semi-supervised learning
we are given a few labeled examples and must make what we can of a large collection of unlabeled examples
Universal Instantiation
we can infer any sentence obtained by substituting a ground term for the variable
knowledge level
we need specify only what the agent knows and what its goals are, in order to fix its behavior
ontological commitment
what a language assumes about the nature of reality
image
what is created by image sensors
environment
what is perceived by the agent
solution
what is returned by a search algorithm with an input, takes form of action sequences when relates to search algorithm
background knowledge
what the KB initially contains
random variables
what variables are called in probability theory, start with capital letters
uncertainty
when an agement doesn't know for certain what state it's in or where it will end up after a sequence of actions
multivalued attributes
when an attribute has many possible values, the information gain measured gives an inappropriate indication of the attribute's usefulness.
optical flow
when an object in a video is moving or the came is moving relative to an object
conditional effect
when condition:effect
belief revision
when inferred facts turn out to be wrong and will have to be retracted in the face of a new information
regression
when the output y is a number, the learning problem is called a regression
classification
when the output y is one of a finite set of values, the learning problem is called classification
nonmonotonicity
when the set of beliefs does not grow monotonically over time as new evidence arrives
loopy path
when you loop in path to a solution. usually bad.
forward-checking CSP
whenever a variable is assigned, the forward-checking process establishes arc consistency for it, ie for each unassigned variable Y that is connected to X by a constraint, delete from Y's domain any value that is inconsistent with the value chosen for X.
scene
where image sensors gather light
intentionality
whether the machine's purported beliefs, desires, and other representations are actually "about" something in the real world
consistency
whether the membership criteria are logically satisfiable
open-loop
while the agent is executing the solution sequence it ignores its percepts because it knows what they will be. because it executes with its eyes closed, it must be certain of what is going on. this is called open-loop because ignoring the percepts breaks the loop between agent and environment
localization
working out where you are given a map of the world and a sequence of percepts and actions
unit resolution
you have a list of things or'd together and'd with a complementary literal of something in the list, then now you have a list without that literal
