1) Search

a) components define [so , Actions G]

3 properly ,
-

So , state space ,
DS-good rep/fu-char/specification
b) general search algo operators , Goal ,

c) evaluation (4) TSCO Path cost

2) uninformed search
a) BFS
y9/ya + 1

>
-
shallowest unexpanded -

FIFoquere -

goal test & gen-memory time

>
-

complete (finite depth) /optimal (equal edge costs)

*
(( /t]
b) uniform
y|
+
cost
>
-

priority guere-lowest cost node-goal dexp

>
-

complete <firits ,
no 0
loops) / optimal (no-ve cost) ba/bd
c) DFS
*
>
deepest 1150 quere- Avoiding
-
-

in
>
-

graph-complete (finite states) / not optimal repeated

tree =
not complete States
d) DIS bel be
>
-

complete incomplete /non optimal

69/bd
-

2) IDDFS
>
-

complete (finite b) /optimal <pathcost is nondes on of depth)

f) Bidirectional search d/2
b
backward search space requirement
>
-

BFS ULS DES PLS IDDS Bid

C yf
ba t
N
bi yFTBs
2
I bM be bd
BFS
y
+
g ys ys
Y Nv
exponential in (relative error in h x
length of solution
Safety Transparency Autonomy Fairness Privacy Accountability
, , , , ,
-
Ethics

3) Informed search
>
-
eval fu >
-
heuristic function h(n)

a) Best first search

>
-
eval In-sorts the queve-cost estimate -
lowest eval

b) GBFS yo
>
-
f(n) =
h(n) ,
fast
1) bd
*
>
-

Incomplete/nonoptimal :
unbounded search space A
 c) A * search
>
-

f(n)
=

g(n) + h(n) admissible ,

neuristic pathmax
>
-
consistency monotonicity : -

f((((f(F) : f(c) =
Max ( + (P) , g(c) + h(c)
>
-

complete/optimal strue cost

>
-
subexponential growth :
In(n)-no(n)/[oCloghW (ul)

d) IDA *

-seauch & within bound ,

increase gradually
>
-
complete/optimal
>
-
time :
dependent on no of values h(n) can take
to
space proportional largest patha
:

5) Beam Search .0 ·

heuristic on performance relaxedons

>
-
-

4) Iterative Improvement algorithms

a) Hill
climbing
no BT-Max on -no· of
>

guere-greedy cost
-
-

no moves
shoulder
> local min plateau ridge
-
, ,
,

>
-
random restart-stochastic -

SA

b) Simulated annealing
annealing
>
-

So Player(s) Action(s) ,
>
-
SECOV, JEJO : D = -Sin ↑ThE
, ,

Result(s , a) Terminal-Test Cs),

,

>
-

reduction schedule Utility (s , P)

5) (SP
6) Game
constraints binary Ho absuls pret
Playing
:
>
-
unay ,
, ,

forward
>
backtracking checking deterministic vls chance
-

>
-

>
-
valiable selection vls
perfect imperfect into

I
Most constrained
> >
challenges :
State space complexity
- -

>
-
most constraining reducing branching -
time constraints ,

modeling
>
-
least constraining vls freedom of future opponent ,
>
-
X : variable
choices
handling uncertainty
a) MINIMAX bM/ bm
D : domains
>
-
min/max (successor's utility
c : constraints
>
-

unification
>
-

complete <finite)/optimal
Copponent)
b) X- B
o is a unifies if EO =
Ez ...
=
EKO GE -EK3 - =

unifiable
pruning _ min : inc
ordering nodes -maxide
.

is of
· a most general unities
c) Expectiminmax
average of leaves
first mismatch of
MINMAX
Disagreement symbols
>
-
.
set :
Dof W is the &

Algorithm : K = 0 WK = W ok = E
, ,

16 Wi is a
singleton stop , ,
return -K

Ele find DK .
16 UK
,
KED & Uk =
van not in tk
substitute
k +
+
7) Logic
logics -
Epropositional ,
Fo Temporal,
,
>
Rep Hypo rept
Fuzzy 3
reason For
-
:

a) Logic Prob ,

To

+
logical power rig math foundath
,
c) k Systems
slow search rigid
-

Ddep KB
,
>
-

Dindp :
IE ,
:

b) Frames >
-
facts , rules , declarative

supports defaults procedural attaches

+
,
>
-

syntax semantics Intrp, inf

, ,

-sound
weak logical power > Inf procedure -
-

complete
-

8) Propositional logic no del

>
-
True : under (intip 16 evals to + CI satisfies #) MP : F+ G
,
F = G

valid : i under all intop E

off TG =
UR : FVC ,

Incon : F under all in tip R :

FVG , TG UH =
FUH
>
-

equivalence : same truth values

A+ B , B + C = A- C
L Act B =
(A + B) 1(B- A)

9) Normal forms (TArB)CIBUA)

DM :
TCFUG) =
TFRTG
-

>
-
Clause : LVLVL for CNF atoms
Terms :
L1L1L for DNF atoms Caveats : False -G =
>
-
Horn clause : less than 1 non-negated literal
>
- FIRF21 .. -
1Fn >
-
G Theorem
F( BC
aX/pos/prm T- T concl , ,

RLNF
>
-
3 proofs :
table , Prm- > Conc Valid ,
Prm1Tconc incon

10) FOL
>
-
Predicate , function FF var ,
const
, ,

VP =
757P
a) Canonical
form JP :
TFTP
>
-

Prenex (FF .... (

>
-
CNF (Matrix)
>
Skolemization (remove])
-

b) domain Interpretation [C , Fn
,
PS
,
Evaluation
,

DDD DITE

c) Scope , Bound , Free

>
-
consistent :
interpretation exists for set of formulas where all one

true
Valid formula : is true every possible interpretation & domain
in

Inconsistent : no interpretation where all in the set are true

simultaneously
Invalid :
at least 1 interpretation of a formula where it is

False
>
-
substitution
0: GV/t 1, V2/ ... 3 to E
EO =
new expression after replacement
composition : 0 . X :
Sul +, 3 .

E + /yz , /y13 :
& "ye "
ly ,
3
,

fu IDA * (problem)
Checkers -

1952 Samuel-10kB
-

roof < Make -

node (Initial-State (problem)
f limit f-cost croot) Backg-1992 -
TesarroNn

LOOP Chess-1997-IBM deep blue

defeated GK
solution , f-limit DFS Contone (root , -

f-limit)
Othello-Reversi-smalless -
if solution ! NUL then return solution
SH performance
it f-limit = = - then return failure
END LOOP x cuts : A
val
In ContoutCroot , f-limit)
75
DFS -
: returns solution , t-limit

:
m/2
Find solution from node root, within the f-cost limit of folimit B cuts Vak B = b

DFS & f-cost limit 11)

Contous returns solution sequence new
Philosophy
-

a) Dualism : M4 B

it G-costCroot) null & f-costroto b) Monism :

f-limitreturn
>
,
solution only 1
if root is solution & its F-cost Crook) c) Idealism M :
a
goal node ,
re

for successors of root do a) Materialism : B

(solution , new-f-limit) < DFS -

Contou (succesors , f-limit

e) Epiphenomenalism : M/B
it new-f-limit is minimum so fal then 12) Psychology
best-so-fact (solution ,
new-f-limit) a) Behaviour

end for b) Functionalism

actua (solution f-limit) of ,

the best-so-far Perpetual
d) Cognitive
① 1913 warren McCulloch , Walter Pitts-NN
1950 Harvard-Marvin Dean-NNCOMP SNARC # 90s-
1956 Darthmouth-birth-logic theorist Newellsimon
HMM , Bayes , SUMIGame Playing
IBM Deep Blue
② 1959- Hubert -
Geom Theory Prove NLP, DMA ,

checkers Chess-Gally
1952-Arthurs progs for play ⑤
-
-

2000
1958 LISP-mccarthy
-

unstruct data
1963 Stanford Allab Big data-info retral
-

1963 James SAINT-closed form

-

calc)probs HLAl , AGI

1968 Tom Evans-ANALOGY-geo
-

analogy 2007
-
holes in pic
1967 -
Danielbobrow -

STUDENT-algebra
③ 60s70s hand code
ELIZA pattern match
>
-
5thgenAl
Prolog
Geneticalgo >
-
NN
80s Symbolic :

logic based