review

Q1: Convert
B

（
P
∨
Q
）

to CNF?

13.1
-

13.10
；

13.13

14.1
（
no e
）
; 14.2.a;

7.1
-
7.2; 7.5; 7.7;

4.3; 4.5;

3.1, 3.6
-
3.7




Introduction:



F
our types of AI



Turing test:
怎么做的，测试什么

acting humanly



Some AI applications, intelligent systems


Agent:



Agent and Ration
ality, performance measure not unique



PEAS



Environment types:



F
ull observable



Deterministic



S
tatic



D
iscrete



S
ingle agent


Problem
-
solving agent:



Problem types:



Deterministic, fully observable


single
-
state problem



Non
-
observable


sensorless problem




No
ndeterministic and/or partially observable


contingency
problem



Unknown state space


exploration problem



Problems:



Romania



Vacuum cleaner



8
-
puzzle



Define a problem



I
nitial state



A
ctions or successor function



G
oal test



P
ath cost



Search



Tree search
：
generate successors of already
-
explored states



N
ode
is a date structure containing much more information
than
state



Search strategy: how to pick the order of node expansion



Properties of search strategy



C
ompleteness



O
ptimality



Uninformed search: only the
information in problem
definition
, no heuristics



Breadth
-
first search
: shallowest unexpanded node



Uniform
-
cost search
: least
-
path
-
cost unexpanded node



Depth
-
first search
: deepest unexpanded node



Informed search




B
est
-
first search with an evaluation functio
n (e.g.
straight
-
line distance) for each node



G
reedy search



A
*

search
:



Local search: hill
-
climbing









Logic agent:



Concepts:
Knowledge base
, syntax, semantics, entailment, models,



Entailment validation:



Check the subset relation



Inference: soundness

and completeness



P
ropositional logic: negation, conjunction, disjunction,
implication, biconditional



Model checking: The truth table for inference, namely
inference by enumeration



Application of inference rules:



L
ogic equivalent: Morgan, distributivity
…



I
nference in Horn form: forward and backward chain



Inference in CNF form: resolution
Question Q1











Probabilistic

agent:



Concepts:

sample point, some space,
probability

space, event,
random variable, syntax for probability, prior probability, posteri
or
probability,



Simple Computation:



Inference by enumeration: compute from the joint distribution
table



P(cavity

j

toothache)



Chain rule



Inference by enumeration: P(Y|E)



Modelling and Computation by conditional independence



Independence



C
onditional indep
endence



Three equivalent definition of CI.



Interprete CI



Modelling joint distribution using CIs



Why CIs make things simpler



A more systematic model: Bayesian Networks



Definition of BNs: syntax and semantics



Compactness



C
onstruction of BNs



Computation:



Bayesian Rule