hobabi lity and Slaisheg

five is drauan tm
a

2) A random samle (hi,kh, y ke,x) of. sige

We kn ow

Nouw

:. t is an unbiased estmado r ot u.

Wen kno
drown Same popuwaton.

2
2 2
 244 4 t

To fnd i 2 such tat t is unbiased.

ano
also

3 3 3

d) best eshinator?

Va, (t) =a
Var) Varls) +Var)+ Vox &u) t Var&r -
2 $

4 Vorki) t
 From the abre Vaionte ,
the best estmator.

Bernoul Vaiate
ue Lf Xi, a. Kn are andom observation s on
the value I with probability p and the value
U oith
a consistent
probabilty l-p. Shou that:
eshrnator of p-p).

P(x)

E(k) = p + o( -p)= P
P
£ (2) tp t o(1 -p)

Var (x) = Pp2

As x folbos Bernaull distibuhon

det
(I-p)
E(6) = mþ : var(6) = np

Take
 lo show
Consistenl entialor of foli)

E(a) ond Von (a) - 0 a4

Von (a) Van ( 2 )

Thus, G is Consis tent estimador

Consider a
Contin wous unchon o*.
°k onsgtent estimator
Thus, by inarient pobey
pllp).
G((- a) is consistent estinador of
Unbiased esimator
Qu3 Show that E Ex-)
n (n-))
, , X,.- 2n dro on on X whch taka the
fhe sample
Tesþecve probabi lie s 6 and (i - 6) .
value with
Kn takes values 0 or L only
Soli Güven
us bernoud? distibuhio rmeany
K X--.. n follo
nean þer &vaione = þ(-) or 6(- 9)
 Nouo for thes

[ng -) n8-n6)

Variance (H) - £(H)- Eu

n tl-g)tn'g = e(')
Now ad hene
£E(Exi-i) unbiaed esimator 6*.
n h-1)

aug Exp lin Gradient descnt Agon ihm, Localy agntd Reyesion
and
So
) avadient descent Atgot thm. t a line to data
to
Gradient descent Aqoi thm is used
and
fhe "o þtimal value for the Rntereþt
slope.
Nouo, that opimal
value is faund by minuni_ing fund

funcin
 rae Size o seþs taken fowahdy minimonm
closer to the min

Value

.Neuo\Value ol lalee- Slepe Leasniy rate

8 = 6 -

Rehaut ses 2 until convajen e

(h,c9 -) (ha-y)
- (hol9-)x;, since, h,(-r

Theetoe6j=-« lhyl*)-r)x;]
inear Regrauisa is geneval ly wed or
the es imahön o lahion ship blu
too paamahers
But, on t be useo or preli'ons
where there exis t a non- (in ea

relaton ship. behsen Kadr.

To non data
orerome the problem o
’ LWR alpithm asigns welghis to data ,
asigned by the hadt of kanal Snuatig
geneal the weight un chion& oe aume d as
()
Heve
inpud thainiy vol
pdy

p h ' I is smal then

egresion ?
I! is Used uhen the responge Vaab le
youab le with tuwo poss b le ralney.
is cakegodal
 Tuwo Succen ond Fai luve,
(4) Co)

ay

Sa) =

Ihy funhon mots ral nodwtth rooabilitiey

is log' t ohi ch is
)ppab abili

0ddy
Colds o sanple
L no. o e v o o me
samle sy