Methods

***
Multivariate

M
~ 0 ALAA DDD
measurements are taken from each point
=> multiple T

tIR
D
*i [xie x=2 XiD]
4th feature
- - -

=
Xi
4) second
fre
Feature feature
(output) value
lebel yi =
target
2 25 21
class
gif D
2 k3
yi)3 EIR
,
- - -

, ,

x E
yi EIR
= ,

motrit

---
S

I
data

·
Xe X12 ----

NXD
X2 ↑

***
N

L L ***
XXX

XXX

XXX

e
220
01270
22 O
20 22170

univerate = XNN(x ; N, 2)
N(
S
Coverence of X, 8 x2 multivariate

&
coverience
mean vector ↓ matrix

= x
= =
 N(x; 0 , 2)
=

E exp[-]
ex) -)
N, )
=
when

=
So y
D=

= ex
Multivariat
Parametric Classificatio
-
class conditionas
~N(
i
1

p(y c)
---

L
=

i ⑧ O
S
)
-

i X I ex
-
Prob
10g[pr(y c)]
[B(xa)
X (gc(x) log(p(x(y c)]
=

= -
EXX + TBa-LaTBa =
= +

~
+ -I leg[pY(y ]
gc(x) = - log(2) Elog(IcD-
+ =

Petel # of perometers =
D +
K(PH) + ·
mean
Vectors
me
nece
matrices
probabilities ba
 = (xi - )
-of data poatsas
)
[i -) 1(yi=
= . .

Om
=
recy
see
of

Pr(y )
= =

)
xWc x + wcX + W
gc(x)
-Y
.

= -

(2)
-
+ leg[P(y c)]
=
-en(x)
ge(x)
xT Wex + w x + wo

we n w 3
=
.

the maximum
= X Nex + + pick
one
"
i
X Wex wa x + wi
g((x) = .
+

-m
-
when T
k = 2
wz)x + (we wa) X + (10 -

220)
xT(we
-
.

gz(x)
-

ga(x)
~
- = ~
-
- Wo
g(x) W W

it to first class
if g(x) 70 >
-

essyn
it to second class.
= assyn
if g(x) <o